a model for calculating interconnection costs in ... · a model for calculating interconnection...

A Model for CalculatingInterconnection Costs inTelecommunications

1818 H Street, NW

Washington, DC 20433

U.S.A.

Telephone: 202.473.1000

Fax: 202.477.6391

Web Site: www.worldbank.org

E-mail: [email protected]

PPIAF Program Management Unit

c/o The World Bank

1818 H Street, NW, MSN I9-907


U.S.A.

Telephone: 202.458.5588

Fax: 202.522.7466

Web Site: www.ppiaf.org


™xHSKIMBy356715zv,:&:':=:"ISBN 0-8213-5671-2

THE WORLD BANK

PPIAF

PPIAF

THE WORLD BANK

A M

od

el fo

r Calc

ula

ting

Inte

rco

nn

ectio

n C

osts

in T

ele

co

mm

un

icatio

ns

TH

E W

OR

LD

BA

NK

The liberalization of the telecommunications markets in Sub-Saharan Africa led toincreased competition on the provision and pricing of communication services. But,

due to the lack of appropriate regulatory tools, newly established regulators are poorlyequipped to arbitrate increasing interconnection disputes between competing operators.This guidebook and its associated CD-ROM, including the cost model, were prepared toprovide Sub-Saharan Africa regulators and operators with a sound regulatory tool allowingthe determination of accurate interconnection costs, thus facilitating the settlement oflengthy and costly interconnection disputes between fixed and mobile operators. The costmodel belongs to the family of “Bottom-Up” models, which calculate interconnectioncost incurred by an efficient operator using the Long Run Incremental Cost (LRIC)methodology. The proposed cost model takes into account most features characterizingthe development stage of telecommunications networks in Sub-Saharan Africa (smallsize of fixed network, importance of rural telephony, excessive reliance on microwavetechnology, explosive demand for mobile service, and weak regulatory capacity).

Paul Noumba Um

Laurent Gille

Lucile Simon

Christophe Rudelle

A Model for CalculatingInterconnection Costs in

Telecommunications

A Model for CalculatingInterconnection Costs in

Telecommunications

Guidebook prepared by Paul Noumba Um, Laurent Gille,Lucile Simon, and Christophe Rudelle

THE WORLD BANK PPIAF

This guidebook is a publication of the Public–Private Infrastructure AdvisoryFacility (PPIAF). PPIAF is a multidonor technical assistance facility, aimed athelping developing countries improve the quality of their infrastructurethrough private sector involvement. For more information on the facility, seethe website at www.ppiaf.org.

The findings, interpretations, and conclusions expressed in this guide areentirely those of the authors and should not, in any manner, be attributed tothe PPIAF, the World Bank or its affiliated organizations, or to members of itsBoard of Executive Directors or the countries they represent. Neither PPIAFnor the World Bank guarantees the accuracy of the data included in this pub-lication or accepts responsibility for any consequence of their use. Theboundaries, colors, denominations, and other information shown on any mapin this report do not imply, on the part of PPIAF or the World Bank Group,any judgment on the legal status of any territory, or the endorsement oracceptance of such boundaries.

The material in this publication is copyrighted. The World Bank holdscopyright on behalf of both the PPIAF and itself. Dissemination of this workis encouraged and, in such instances, the World Bank will grant permissionpromptly, and will not require a fee when reproduction is for noncommercialpurposes. Any request to photocopy portions of this publication should beaddressed to: Copyright Clearance Center, Inc., 222 Rosewood Drive, Dan-vers, MA 01923, U.S.A., telephone 978-750-8400, fax 978-750-4470, orthrough the Internet at www.copyright.com.

In regard to questions concerning this publication, or information onordering more copies, please refer to the PPIAF website, or contact PPIAF byemail at the address below.

PPIAFc/o The World Bank1818 H StreetWashington, DC 20433Fax: 202-522-7466www.ppiaf.org Email: [email protected]

ISBN 0-8213-5671-2

Library of Congress Cataloging-in-Publication Data

Modèle de détermination des tarifs d’interconnexion. English.A model for calculating interconnection costs in telecommunications /

edited by … [Paul Noumba Um].p. cm.

Includes bibliographical references.ISBN 0-8213-5671-2

1.Telecommunications—Rates—Africa, Sub-Saharan—Mathematicalmodels. 2.Telecommunications—Africa, Sub-Saharan—Costs—Mathematical models. I. Noumba Um, Paul. II.Title.

HE8467.M6313 2003384.6'4—dc22

2003067270

Copyright © 2004 International Bank for Reconstruction and

DevelopmentTHE WORLD BANK

1818 H Street, NWWashington, DC 20433, USA

Telephone 202-473-1000Internet www.worldbank.org

E-mail [email protected]

All rights reservedPrinted in the United States of America

First printing: December 20031 2 3 4 07 06 05 04

v

Foreword viiiAcronyms and Abbreviations ixAbout the Authors x1. Introduction 1

Preliminary Results Provided by the Cost Model 2The Limits of the Model 4

2. Guidebook Overview 6Cost Modeling Principles 6User Guide 8The Model 9

3. Cost Modeling Principles 11Regulation of Interconnection 11

Essential Facilities 11Cost Determination Methodologies 14Economic Criteria for Assessing Costs 15The LRIC Concept 18Advantages and Drawbacks of LRIC 21Definition of Interconnection Services 22Definition of the Increment 23

Application to African Telecommunications Networks 25The Average Size of an African Telecommunications Network 25Nodes and Links 26Transit 27

4. Modeling PrinciplesThe Modeling Principles 29

A Long-Run Approach 29A Forward-Looking Approach 29An Efficient Approach 29An Economic Approach, Not an Accounting Approach 30A Bottom-Up Approach 30The Working Units 30

The Model Structure 30Applications of the Cost Model 32

Contents

vi

5. User Guide 34Filling in the Assumptions 35

Demand Assumptions 35Main Technical Assumptions 38Main Assumptions on Routing Factors 41Main Cost Allocation Assumptions 42Assumptions for the Mobile Network 45

The Results Sheets 47The Results 47Printing Reports 49Testing the Sensitivity of the Results 49Managing the Model 51

6. Operations of the Cost Model 52Basic Principles 52Logic of the Intermediary Spreadsheets 53

Capacities (Capa Sheet) 53Transmission Capacities (Capa ElTr Sheet) 53Infrastructure Capacities (Capa Infra Sheet) 54Costs 57

The Mobile Network Calculations 58Capacity of the Mobile Network (CapaMob Sheet) 58Cost of the Mobile Network (Costs Mob Sheet) 58Total Costs of the Mobile Network (Tot Mob Sheet) 59

Appendixes 60Appendix 1: Economic Cost Approach 60Appendix 2: Capital Cost Approach 63Appendix 3: Radio Concentrator Solutions 66

Bibliography 67Tables

1.1 A Sample of Interconnection Rates 11.2 Overview of the Telecommunications Sectors in Burkina Faso, Côte d’Ivoire, Cameroon,

and Zambia 31.3 A Sample of Interconnection Rates Calculated by the Cost Model 32.1 Typology of Nodes and Links in a Telecommunications Network 73.1 Correlation Matrix between Retail Services and Elementary Services 233.2 Matrix of Inputs Needed to Provide Retail Services 243.3 Snapshot of Telecommunications Development in Africa in 1999 253.4 Matrix of Transmission Links 27

A3.1 Number of Subscribers Depending on the Traffic per Subscriber (in mE) 66

Figures1.1 Comparison between Interconnection Prices: Côte d’Ivoire 41.2 Comparison between Interconnection Prices: Cameroon 42.1 Cost Model Architecture 103.1 The Costs of a Multiproduct Firm: An Example for Five Products 143.2 Transition from Historical Accounting Costs to Economic Costs (LRIC) 193.3 Cost Structure 203.4 Entry Points of Interconnection Service 234.1 Cost Model Architecture 326.1 Cost Model Architecture 52

A3.1 Example of the Architecture of an IRT (TRT-Lucent Network) 66

Contents

Boxes3.1 Different Definitions Selected for Cross-Subsidization 175.1 Erlang 396.1 Sizing of the SDH Network (example of RCU-LS routes) 54

Contents

vii

viii

The tremendous economic and social impactresulting from pervasive, as well as effective, use ofinformation and communication technologies hasconvinced more and more African governments toembark on the restructuring and progressive liberaliza-tion of their telecommunications sectors.

In a liberalized market, a procompetitive regulatoryframework is a factor in establishing a level playingfield for a favorable environment that encouragesgreater participation by the private sector. In thatrespect, establishing a sound interconnection frame-work that ensures equal treatment, nondiscriminationamong market players, and cost-based tariffs is assumedto be the main engine for developing competition inthe sector.

Determining interconnection tariffs is a complexand extremely sensitive task. In Africa, the absence ofaccurate cost information has rendered the situation allthe more complex. In fact, some telecommunicationsregulators resolve interconnection disputes on thegrounds of available tariff benchmarking, althoughthese tariffs may not always be relevant.

This guidebook provides a sound methodology tohelp regulators and telecommunications operatorsadopt a tariff regime and deal with interconnectiondisputes on the basis of a rigorous cost model.

Furthermore, it demonstrates the World BankGroup commitment to assist governments in develop-ing the proper enabling environments and effectiveregulatory institutions, and hence, contribute effec-tively in mitigating the regulatory risk as perceived bypotential investors. Helping newly established regula-tory agencies develop their capacity has been under-scored as a key pillar of the new World Bank GroupInformation and Communication Technologies (ICT)strategy (www.worldbank.org/ict).

We sincerely hope that this guidebook and its costmodel will be used by telecom regulators and opera-tors to settle interconnection disputes. However, wewould urge regulators to customize and expand thecost model to better match their internal needs.

The guidebook consists of six chapters. Chapter 1introduces the guidebook. Chapter 2 provides anoverview of the guidebook. Chapter 3 reviews costmodeling principles for calculating interconnectionrates, while chapter 4 illustrates how these principlesare integrated in the cost model. Chapter 5 is a userguide and chapter 6 reviews cost modules composingthe model and illustrates how interconnection costs arecomputed.

Mohsen A. Khalil

Foreword

ix

ABC Activity-based accountingADM Add-drop multiplexerBHCA Business hour call attemptsBHE Business hour ErlangsBSC Base station controllerBTS Base terminal stationCS Central stationECPR Efficient component pricing ruleEU European UnionFDC Fully distributed costsGDP Gross domestic productGICT Global Information &

Communication Technologies IS International switchIT Information technologyLRAIC Long run average incremental costsLRIC Long run incremental costLS Local switch

MSC Mobile switching centerMUX MultiplexerNRA National regulatory authoritiesPDH Plesiochronous digital hierarchyPPIAF Public–Private Infrastructure

Advisory FacilityRCU Remote concentrator unitRIO Reference interconnection offerSDH Synchronous digital hierarchyTDMA Time division multiplexing accessTELRIC Total element LRICTRX Transmitter/receiverTS Terminal stationTSLRIC Total service LRICTSW Transit switchUSO Universal service obligationWACC Weighted average cost of capitalWBI World Bank Institute

Acronyms and Abbreviations

Paul Noumba Um is lead infrastructure specialist,World Bank Institute (WBI),The World Bank Group.Previously, he was regional coordinator for Africa inthe policy division of the Communications and Infor-mation Technology department.

Laurent Gille is head of the department of economicsand social science at École Nationale Supérieure des

Télécommunications (ENST) Paris, 46 Rue Barrault,Paris 75013 France. Previously, he was head of the reg-ulation group in BIPE SA.

Lucile Simon and Christophe Rudelle are consultantswith BIPA SA – L’Atrium, 6 place Abel Gance,Boulogne Billancourt, 92652 France.

x

About the Authors

1

Since the past decade, several Sub-Saharan Africangovernments, through technical assistance provided bythe World Bank and other donors, have undertaken toreform their telecommunications sectors, by imple-menting market liberalization policies, privatizing theincumbent public operator, and creating autonomousand independent regulatory bodies.The core objectiveof these reforms is to significantly improve access, andaffordability, to telecommunications services on thebasis of the assumption that a more friendly and pre-dictable business environment will attract more privateinvestment. However, the provision of interconnectionservices, on fair and efficient terms, has rapidlyemerged as a main bottleneck.

In fact, new legislation and regulations enacted inSub-Saharan Africa recognize the interconnectionrights ascribed to all telecommunications serviceproviders and network operators. In addition, theseregulations also request the incumbent fixed operatorto supply interconnection services to new entrants on

a fair and competitive basis. Despite the clarity andsoundness of the legislative provisions in that respect(cost oriented, nondiscriminatory, fair, and transpar-ent), the number of interconnection disputes hasincreased, and long-lasting interconnection disputeshave discredited the reputation and credibility of newregulatory regimes.

One has to admit, however, that deriving optimalinterconnection rates from the principles codified innational laws is tricky in countries constrained byscarce resources (human resources with relevant expe-rience, management and information systems that areunable to provide accurate and comprehensive data,and so forth). As illustrated in table 1.1, interconnec-tion rates are mostly derived from international bench-marking. Although the international benchmarkingapproach may be a satisfactory starting point, it is notalways relevant, since it may not take into account spe-cific country parameters that affect the industry coststructure.

1 Introduction

Table 1.1 A Sample of Interconnection Rates

Euro cents/minute Benin Burkina Faso Burundi Cameroon Côte d’Ivoire Mali Mauritania Togo

Local 4.5 3.5 3.8 4.0 3.8 3.8 2.6 6.1Simple transit 4.6 27.8 3.8 8.6 19.8 16.4 7.6 6.1Double transit 19.8 27.8 3.8 19.8 19.8 16.4 7.6 6.1Transit 1.8 1.2 3.8 2.3Mobile termination 19.8 13.4 3.8 22.1 23.7 30.0 7.6 9.9Source: Data collected from participants of the Access Pricing Workshop held in Ouagadoudou, Burkina Faso, in March 2002.

A Model for Calculating Interconnection Costs in Telecommunications

2

At the beginning of the sector reform in Africa, inthe mid-1990s, it was believed that incumbent fixedoperators would hold their dominant position in thelong run. However, the recent explosive developmentof cellular business substantially contradicted that pre-diction. In most countries, mobile operators connectmore subscribers than the fixed service. In a short time,cellular operators have achieved significant marketpower and are becoming dominant players themselves.Therefore, regulation of interconnection and accesspricing should not only focus on calls terminating inthe fixed network, but also should consider calls origi-nating or terminating in the mobile networks.

The World Bank Global Information & Communi-cation Technologies (GICT) Department believes thatdeveloping a more accurate and robust methodology toassess interconnection rates, based on long-term incre-mental costs, will support efforts by borrower countriesto implement best practice regulation, and would sig-nificantly improve the reputation and credibility ofnational regulatory authorities (NRAs). Our hope is tosee this generic cost model customized and expandedby NRAs and be used to solve pervasive interconnec-tion disputes, which have flourished across Africa.

The guidebook includes a CD-ROM that containsthe bottom-up cost model.The cost model was devel-oped by BIPE SA1 under a Public–Private InfrastructureAdvisory Facility (PPIAF) grant managed by the WorldBank. It builds, to some extent, on a model initiallydeveloped by Europe Economics,2 at the request of theEuropean Commission. However, the proposed modeltakes into account the specific features characterizingtelecommunications development in Africa (embryonicsize of the network, predominance of microwave tech-nology for transmission links and limited roll out offiber-optic cables, rollout of expensive time divisionmultiple access [TDMA] systems to connect rural local-ities, and limited regulatory capacity), and calculatesrouting factors, in the light of parameters entered by theuser. It generates interconnection rates for fixed-to-fixed, fixed-to-mobile, and mobile-to-fixed calls. Othervalue-added services are not captured in the attachedversion but could be easily added by users.

The cost model is available at www.worldbank.org/cit and the Europe Economics model3 isavailable at www://europa.eu.int/ISPO/infosoc/telecompolicy/en/Study-en.htm.

The guidebook was prepared by a team led by PaulNoumba Um (World Bank), including Laurent Gille,Lucile Simon, and Christophe Rudelle from BIPE SA(France).The team made use of comments and guid-ance provided by Antonio Estache and David Satola(World Bank). Daniel Benitez (IDEI, University ofToulouse) reviewed the cost model and provided sug-gestions for its improvement. We gratefully acknowl-edge valuable comments provided by participants tothe Interconnection Day, which was organized by theGICT Department on April 10, 2003.

We equally wish to express our gratitude for thevaluable support offered, throughout this project, byMichele Rajaobelina, Lizmara Kirchner, and LucyCueille (World Bank). This exercise could not havebeen completed without the meaningful collaborationand support provided by telecommunications opera-tors and regulators in the following countries: BurkinaFaso, Côte d’Ivoire, Cameroon, Senegal, and Zambia.KC Translations Services LLC revised the English ver-sion of the guidebook.

Preliminary Results Provided by the CostModel

From July 2002 to February 2003, the World BankGICT Department commissioned BIPE to conduct fieldvisits to Burkina Faso, Cameroon, Côte d’Ivoire, andZambia. During these visits, workshops were conductedto train regulator and telecom operator staff on how touse the cost model.This section provides an overview ofthe preliminary findings derived from these visits.

The four countries have implemented telecommu-nications reforms and established sound regulatoryframeworks enabling competition in specified marketsegments such as the mobile market. Although thesecountries are not homogeneous, they do, however,reflect a series of commonalities in terms of regulatoryframework, market structure, and overall telecommu-nications development. Interconnection disputes andthe urgent need for regulators to ensure their effectiveand timely settlement were identified as among theprimary concerns of private mobile operators.The fol-lowing table provides an overview of the telecommu-nications sector situation in these countries.

With the exception of Cameroon, and Côted’Ivoire, fixed incumbent operators have been granted

Introduction

3

mobile licenses. In general, the dominant position offixed incumbent operators is seriously challenged bythe explosive growth observed in the mobile marketsegment.

In the four countries, interconnection rates werefinally decided by regulators in an attempt to settlelengthy interconnection negotiations. In Zambia,interconnection disputes were subsequently brought tothe law courts by mobile operators, and are still not set-tled. In Côte d’Ivoire, following the submission of therevised interconnection rates in 2000 by the fixedincumbent, mobile operators filed complains to thelocal regulator (ATCI). The arbitration published in

2001 by ATCI was appealed by the fixed incumbent toCTCI, the sector’s appeal court for disputes betweenoperators and ATCI. In November 2002, CTCI pub-lished its decision on interconnection rates. Neitherthe ATCI arbitration nor the CTCI appeal decisionwas based on sound economic analysis.

The preliminary results provided by the cost modelduring these field visits are summarized below. InZambia, the fixed incumbent was unable to providetraffic information required to run a cost simulation. InCôte d’Ivoire, arbitration was decided by CTCI, andresulting rates remain higher than the ones recom-mended after the workshop. In Burkina Faso, the regu-

Table 1.2 Overview of Telecommunications Sectors in Burkina Faso, Côte d’Ivoire, Cameroon, and Zambia

2001 Data Burkina Faso* Côte d’Ivoire Cameroon Zambia

Subscribers to the fixed network 60,000 300,000 105,000 105,000

Subscribers to the mobile network 111,145 730,000 510,000 150,000Number of mobile operators 3 3 2 3* 2002 data.Source: BIPE 2003.

Table 1.3 A Sample of Interconnection Rates Calculated by the Cost Model

Euro Cents Burkina Faso Côte d’Ivoire Cameroon Zambia

Fixed NetworkCurrent Rates

Local 3.1 9.8 4.0 NASimple transit 14.9 9.8 8.5 5Double transit 14.9 19.8 19.8 NA

Cost Model Rates (with TDMA systems included)Local 1.0 3.3 1.7 NASimple transit 1.8 4.4 9.6 NADouble transit 2.4 5.2 12.7 NATransit 0.5 0.8 3.1 NAInternational transit 2.5 1.0 6.4 NA

Mobile NetworkCurrent Rates

Local 9.5 15.2 22.1 5.0Cost Model Rates Celtel Telecel Telmob Orange Telecel Orange MTN Celtel TelecelOriginating 13 9 8 5.7 7.7 27.2 24.4 14.3 11Termination 13 9 8 9.8 9.4 27.2 24.4 14.3 11NA, not applicable.Notes:The interconnection prices for collection and termination are not theoretically symmetrical in so far as this traffic does not call on the same network elementsin a symmetrical way.This lack of symmetry can be taken into account by means of routing factors. However, in the cases of Cameroon and Zambia, we worked withdefault routing factors that do not discriminate between collection and termination traffic, hence the equality of the results found.We indicated to the operators andregulators that they should refine the calculation of the default routing factors in order to better take into account the real use of the various network elements attraffic collection and termination levels. In reality, the notion of transit corresponds to the difference between double transit and single transit, in a case where a thirdparty operator wants to route traffic on the fixed-line operator’s network from one transit zone to another transit zone, while collecting and terminating the trafficitself. For further details, see the model’s user guide.Source: Authors’ own calculations.

lator (ARTEL) enacted an interim regulation on inter-connection rates following a complaint filed by one ofthe mobile operators. During the workshops, the dis-cussions and debates were less emotional and weremore focused on specific issues related to the relevanceof the model assumptions or parameters.

Although it is too early to derive any definitiveconclusion from this experience, the empirical resultsprovided by the cost model are quite robust. In Côted’Ivoire, the interconnection rates generated by themodel were below the rates ratified by the regulator forsingle, double transit, and call termination on themobile networks. For these services, interconnection ischarged above costs, and interconnection providers areextracting monopoly rents. Conversely, local intercon-nection seems to be cost oriented. Additional actionsfrom the local regulator are therefore needed. In con-trast, the simulation in Cameroon provided mixedresults.The incumbent’s network is not optimized, andcosts incurred to provide interconnection services tocompetitors are therefore abnormally high. Finally, theresults show important differences between mobiletermination costs in Côte d’Ivoire and Zambia on theone hand, and Cameroon on the other hand.These dif-ferences can be explained by substantial investmentsmade by mobile operators in Cameroon to bypass theincumbent,which did not have excess capacity to meetthe demands of the mobile operators.

The Limits of the Model

Although the proposed cost model builds on Africantelecommunication network specificities, it can beapplied in non-African environments provided appro-priate adaptations are made.The model is easy to use,and requests information that regulators or operatorscan easily find.However, there remain some limitations.First, the model is designed for “small”networks that donot yet implement complex transit functions or routingalgorithms. Second, the model does not seek full opti-mization when rebuilding the transmission network.More specifically, the model does not optimize thenodes, as the current network topology is keptunchanged (“scorched node” or Brownfield approach).Third, it does not provide a detailed modeling of thecables and duct networks, though it discriminatesamong different types of geography (urban, suburban,rural) or nature of the trenches (wrapped, ducted,buried). Such a module could be developed by eachregulator, although node location optimization may notbe critical at the current stage of telecommunicationsnetwork development in developing countries.

In conclusion, the proposed cost model providesaccurate cost proxy estimates when applied to net-works with less than 1.5 million main lines and whenthe incumbent’s network topology (number and loca-tion of nodes) is optimized or close. As shown in


4

Figure 1.1 Comparison between Interconnection Prices:Côte d’Ivoire

0

5

10

15

20

25

Local Singletransit

Doubletransit

Collection Termination

Euro cents

Current prices Model s results - fixed

Model s results - mobile 2Model s results - mobile 1

Figure 1.2 Comparison between Interconnection Prices:Cameroon

0

5

10

15

20

25

30

Euro cents

Local Singletransit

Doubletransit

Collection Termination

Current prices Model s results - fixed

Model s results - mobile 2Model s results - mobile 1

Cameroon, the results obtained when the incumbent’snetwork is underoptimized may be misleading. In asubsequent version of the cost model, an “earth node”module will be added to generate interconnectioncosts when node location and links are optimized.Therefore, regulators will be able to calculate abounded interval of interconnection rates. The lowerbound will correspond to a configuration of a fullyoptimized network (nodes and links), while the upperbound will correspond to a partially optimized net-work (only links are optimized).

Introduction

5

Notes

1. BIPE SA can be contacted at: L’Atrium, 6,place Abel Gance, F92652 Boulogne Billan-court Cedex,Tel.: 33 (0)1 46 94 45 22,Fax: 33 (0)1 46 94 45 99,E-mail:[email protected], http://www.bipe.fr.2. Europe Economics Research Ltd. (EuropeEconomics) can be contacted at: ChanceryHouse, 53-64 Chancery Lane, London

WC2A 1QU, Tel.: (+44) (0) 20 7831 4717,Fax: (+44) (0) 20 7831 4515.3.See “Final Report on the Study of anAdaptable Model Capable of Calculating theForward-Looking, Long-Run IncrementalCosts of Interconnection Services for EUMember States” (April 2000), prepared forthe European Commission by Europe Eco-

nomics. This study resulted in the produc-tion of a model spreadsheet in MS-Excelformat [EN, 4 Mb] (with a voluminous userguide),which is described in the main report[EN, 440 kb] and an executive summary[EN, 65 kb] (both available here as AdobeAcrobat *.pdf files).

6

2 Guidebook Overview

Interconnection negotiations and settlements areamong the main regulatory issues to reckon with inregard to the development of competition in thetelecommunications sector in African countries.Although the majority of legislation includes provi-sions ensuring the interconnection rights for newentrants and sets cost orientation principles for deter-mining the interconnection rates, only a few regulatorsare equipped to effectively implement these regula-tions in practice. Regulators do not have relevant costinformation that would allow for effective arbitrationof interconnection disputes. Moreover, regulators arenot equipped to assess the cost orientation of intercon-nection rates proposed by fixed incumbents. Hence,they are ill equipped to ratify reference interconnec-tion offers.

Furthermore, it is difficult to replicate cost modelsdeveloped for more advanced economies in Africa.African networks are small in size and quite spread out.They rely on specific architectures and technologies toreconcile their small market size in volume to the scat-tered habitat. Transmission links are mostly overmicrowave technology, and the roll out of fiber-opticsystems remains limited to urban centers.

It is in this context that the World Bank contractedBIPE to develop a cost model that captures specificitiesprevailing in Africa and that could be easily replicable.The model belongs to the bottom-up, long run incre-mental cost (LRIC) models family, and determinesinterconnection rates by calculating the costs incurred

by an efficient African network operator using the bestavailable technologies. Considering the tremendousdevelopment of mobile communications in Africa, themodel also calculates the interconnection cost forfixed-to-mobile calls, and conversely.

In general, the model requires the entry of substan-tial amount of information characterizing the net-works that are interconnecting (topology, architectureprinciples and rules, traffic matrix and patterns, costselements, and so forth).To palliate the frequent gaps inthe information systems, the model proposes defaultvalues that the user could consider if needed. Thisapplies mainly to the routing factors, which are rarelydocumented in most countries.

The overall objective is to provide regulators andoperators with a decision tool that can enhance commonunderstanding and cooperation in dealing with a sensitivesubject.The next sections briefly present the cost mod-eling principles and the user guide and review the costmodel structure.

Cost Modeling Principles

Nobody could envisage dynamic competition in thetelecommunications market without the actualenforcement of the interconnection regulations.With-out interconnection, competing operators will beobliged to duplicate onerous infrastructure, and con-sumers would have to subscribe to connections to dif-ferent operators’ networks.Conversely,with an effective

interconnection regime, a seamless communication sys-tem is more likely to develop, enabling consumers tocontract the service with whichever supplier, and beensured of receiving all incoming calls, from whereverthey originate. Nonetheless, the interconnectionregime has to be implemented with market liberaliza-tion in view; and this calls for free negotiation and con-tracting. Hence, interconnection agreements have to befreely negotiated in line with regulations. That keyprinciple—as it is applied—opens the door to abuseand strategic behavior from the operator enjoying adominant market position. Fixed incumbent operatorsare likely to enjoy a strategic comparative size advantage(number of subscribers) when the market is opened upto competition, which can distort or hamper the devel-opment of competition. Safeguards are, therefore,needed to protect new entrants from these anticompet-itive behaviors; hence, regulators are mandated by lawto ratify reference interconnection offers submitted byincumbents. It is also required that they ratify agree-ments negotiated by the parties. In so doing, regulatorsmust not only make sure that the agreements areentirely in compliance with existing regulations butthey must also check their consistency with the guidingprinciples, such as interconnection rates cost orienta-tion. In checking compliance with the cost-orientationprinciple, regulators need to access relevant and accu-rate cost information to motivate their decisions. How-ever, assessing the cost orientation also requires anability to assess the way costs for interconnection ser-vices are formed and distributed.

This is a complex and cumbersome task. Indeed, atelecommunications operator manages a complexbusiness. It rolls out networks using different technolo-gies and investment layers spread out over time. Itoffers a portfolio of services, which are interdependentand call on the same productive resources. Further-more, some of the services are sold on a retail basis (inthe final market) while others, such as interconnectionservices, are sold to other operators forming a whole-sale range of products. In such a context, assessing oridentifying costs incurred by each category of servicerequires the implementation of a sophisticated cost allo-cation management and information system.

Economists have proposed various cost allocationmethodologies. Some are based on the operator’s

accounting and involve allocating historical costs todifferent services according to criteria prescribed bythe regulators. Others estimate costs by reconstitutingthe networks on the basis of currently available tech-nologies. It is generally admitted that the latter are themost appropriate for estimating interconnection ser-vices costs. LRIC methodology estimates the costsincurred, while offering a subset of services.The costsconsidered are those that would be avoided if theseservices were not offered.

For estimating the cost of the interconnection ser-vices, the selected increment comprises network ele-ments belonging to the core network, that is, thoseshared among all the network’s users, excluding net-work elements dedicated to end users.

The proposed cost model takes into account thespecific nature of African networks and has several fea-tures:• A low number of main lines spread out, however,

over large territories.• Traffic concentrated over a small number of net-

work nodes.• Transit function almost nonexistent, and low capac-

ity of the transmission network,which relies merelyon microwave technology.

• Predominance of rural concentration systems thatuse TDMA-type radio systems.

• The presence of domestic satellite networks.These specific features are taken into account in the

model, and rural radio concentrators are integrated inthe increment.

The cost model takes into account six types ofnodes and five types of links between these nodes, assummarized in the following table:

IS, international switch;TS, terminal station; LS, local switch; RCU, remoteconcentrator unit; CS, central station.

Guidebook Overview

7

Nodes IS TS LS RCU CS TS

ISTS to IS TS-LS TS-LSLS TS-LS RCU-LS (Local RCU link)CS CS-TSTS

Table 2.1 Typology of Nodes and Links in a Telecommunications Network

The cost model assumes the existence of two tran-sit levels: international switch (IS) and domestic transit,which are often not implemented in practice.The tran-sit functions are often performed by the local switches(LSs), which are also used to connect subscribersdirectly or indirectly through remote concentratorunits (RCUs). Finally, TDMA technology radio con-centrators, characterized by central stations (CSs) andterminal stations (TSs), are widely used to connectremote rural localities to the fixed telephone network.

The cost modeling assumes that costs are:• Long-term—meaning that all the cost compo-

nents are variable.• Forward looking—implying that the model con-

siders current costs and not historic costs.• Efficient—this implies that the model takes into

account the best available technology.This is donewithout modifying the network topology. In thiscase, the cost model retains the scorched nodeapproach.

• Economic—not accounting costs. For instance,the model converts investment costs into constantequivalent annuities.

• Bottom-up—this implies rebuilding the networkaccording to the previous principles.

• Computed per minute.The cost model is Excel based and includes 21

spreadsheets, 7 of which can be accessed via an interfaceprovided in the Menu sheet. For fixed networks, themodel calculates interconnection costs for local, singletransit, double transit, and international transit calls. Formobile networks, the model calculates interconnectioncosts for both terminating and originating calls.

User Guide

The user guide is intended to facilitate the use of themodel by staff from regulatory agencies and telecom-munications operators.

After selecting a working language (English orFrench), the user is requested to provide genericparameters defining the network’s configuration, archi-tecture, and topology.

The user enters in the blue boxes parameters char-acterizing the network’s configuration in terms of size,architecture, topology.

The parameters describing functionality shared byseveral networks are filled in by default values.The usercan modify these values whenever needed.Two situa-tions are possible:• Default values are not calculated values and are

filled in light green boxes. In this case, the user canmodify these values.

• The default values are calculated from the user’sinputs and are filled in table format on the right ofblank light blue cells. In case the user wants to pro-ceed with default values, then he/she should avoidfilling in alternative values.Once all of the required information is entered or

validated, the user can view the results presented insubsequent tables.• The first table gives the cost per minute of various

network elements (nodes and links).• The second table gives interconnection services

costs as follows:– The first line provides the average interconnec-

tion costs resulting from selected routing factors.– The second line supplies the interconnection

costs with or without factoring in TDMA systemcosts. These costs are compared with Europeanbest practice rates provided in euros.

– Finally, the table translates average costs obtainedin tariffs according to the pricing structure inforce (depending on time or day).

• Similar simplified tables are provided for cost of ter-minating or originating calls from mobile networks.The user can refine certain assumptions and con-

duct a sensitivity test.This can be done for six prese-lected variables:• Traffic level at peak time (as a percentage of the

overall traffic).• Total length of trenches: variation by percentage.• Proportion of staff dedicated to the core network

functions (maintenance, operation, and so forth).• Employee average annual cost.• Markup to the equipment capital cost incurred by

the operator (this is a proxy reflecting exogenousfactors that increase capital cost).

• The proportion of debt in the total capital structure.The user can print a report including the simula-

tion results. Similarly, the user can save the sensitivitytest modifications.


8

Guidebook Overview

9

The Model

The model is described in greater detail in the thirdsection of this guidebook. The overall logic of themodel is simple.• The model begins with a nomenclature of network

elements (nodes and links).• Each service uses these elements in different pro-

portions.The routing factors represent the averagenumber of times a given element is used by the ser-vice considered.The model then calculates the totalload supported for each network element.

• The model calculates the size of network elements(transmission elements) within the framework ofthe selected topology.

• The model adds up all corresponding network ele-ment costs and calculates the per-minute cost foreach network element.

• Finally, the model calculates the interconnectioncosts on the basis of the routing factors.The network is considered to be made up of net-

work elements, namely, nodes and links. These ele-ments convey interconnection traffic throughout thenetwork.This implies the mobilization of network ele-ments according to the complexity of the interconnec-tion service requested.

Furthermore, the model calculates the investmentcost and also the cost of operation and maintenanceincurred by these network elements. These costs aredistributed over four components.

Common costs are expressed as a percentage of theattributable costs.The percentage is usually decided bythe regulator and is to some extent arbitrary. Theattributable costs are the costs directly caused by theinterconnection service and would have been avoidedby the provider.They consist of operating and invest-ment costs.The operating costs are composed of twoterms:• Maintenance and operating cost related to the net-

work element (spare parts, equipment section of

preventive and corrective maintenance, energy con-sumed).

• Staff costs related to operation and maintenanceactivities.For small but spread-out networks, staff costs can

hardly be appraised as a percentage of the investmentcosts.This is, indeed, a difficult task.The model calcu-lates the number of staff needed to efficiently operatethe network and derives salary costs accordingly. Thestaff cost is then distributed among different networkelements according to parameters specified by the user.

Similarly, the investment costs are calculated byconsidering the estimated volume of traffic to be han-dled by each network element at peak hour.The size ofeach network element is accordingly derived usingengineering procedures. For each network element,the model calculates the investment cost incurred andgenerates an investment annuity. Operating and non-attributable costs are then allocated to it to obtain aglobal cost per minute for each element. For each ser-vice, relevant costs per minute for the network elementare added up to obtain the interconnection cost.Thelatter value is adjusted using the gradient of retail pricesto derive the interconnection rate.

The model calculates the interconnection costsover two stages.• First, the model determines the size of the switching

elements. Switching elements are considered to bethe nodes of the network. Their investment costsdepend on the switching system BHE (business hourErlangs transformed into 2 megabytes per second[Mbps]) and the number of subscribers connected.

• Second, the size of the links connecting network’snodes is calculated. The model differentiates theinfrastructure and transmission layers:– The transmission layer enables the user to design

the electronic transmission equipment by choos-ing and sizing the capacity of the most appropri-ate technology to be rolled out (synchronousdigital hierarchy [SDH] rings, plesiochronousdigital hierarchy [PDH] technology).

– The infrastructure layer enables the determina-tion of the links’ substratum or the physical ele-ments that will support the transmission link:trenches for fiber-optic cables, microwave towersand masts, and satellite earth stations.

Investment Operatingcosts costs

Attributable costsCommon costs

The trenches are broken down by geo-type (urban,suburban, and rural) corresponding to various buryingtechniques (wrapped trench, ducted, fully buried).Themicrowaves are characterized by the nature of theirmass (light, medium, or heavy).

The links are sized by the traffic carried at peakhour expressed in megabit per second (Mbit/s). Thetransmission link size is determined, to enable normalflow of traffic expected from switching centers andleased lines connected to them. Certain infrastructureelements are shared by different categories of traffic;this is particularly applicable to SDH rings in the accessnetwork. The model enables such costs to be shared.For example, SDH rings in the access network will beused to carry core network traffic flow and the accessnetwork traffic.

The model comprises 21 sheets organized as fol-lows:• One menu sheet.• Twelve sheets forming the core of the model, as

described below.• Four sheets for the mobile networks.• One sheet to conduct the sensitivity analysis.• Three specific management sheets (two sheets for

publishing the fixed-line and mobile reports, andone sheet to manage the two languages and thedefault values).The 12 sheets forming the core of the model are:

• Four sheets for assumptions and parametersenabling the configuration of the network.

• One sheet for calculating the peak load traffic anddetermining the network’s elements size.

• Two sheets for determining the transmission andinfrastructure size.

• Three sheets to calculate the network element costs(switching, transmission, and infrastructure).

• One sheet for summing up the total interconnec-tion costs (including the shared costs) and calculat-ing the interconnection costs per minute and perelement.

• One sheet to present the results.Figure 2.1 below summarizes the cost model archi-

tecture.In conclusion, this cost model is a tool provided to

regulators and operators to help them:• Develop a better understanding of the interconnec-

tion costing and economics.• Determine economic-oriented interconnection

rates for terminating and departing traffic from, orto, fixed and mobile networks.

• Ratify crucial assumptions pertaining to trafficdemand, network optimization, and cost allocationbetween final and intermediary services.

• Report on the benchmarking exercise, for the rat-ification of the above mentioned assumptions.Instead of benchmarking interconnection rates, the regula-tor should instead focus on benchmarking key cost driv-ers.

• Better identify indicators to be monitored, on aregular basis, by the regulator.


10

DemandDemand

NetworkRoutingFactors

Traf c

Transmission Infrastructures

Switching Transmission Infrastructures

Unit CostsUCosts

Totals Results

Hypotheses

Traf c

Sizing

Cost

Result

Tech FactRout

Capa Eltr

Costs Sw

tot results

Costs Tr Costs Infra

Capa Intra

Capa

Figure 2.1 Cost Model Architecture

11

Regulation of Interconnection

Since the past decade, the majority of African countrieshave adopted legislation opening up their telecommu-nications market, or certain segments of it, to competi-tion. The mobile communications segment wasopened up to competition, with two to five mobileoperators authorized to commercialize their services.In regard to fixed-line networks, the scope of compe-tition remained limited as temporary exclusivity peri-ods were accordingly granted to incumbent fixedoperators before or after their privatization.

In parallel, independent or autonomous regulatorybodies were established in the majority of Africancountries with the mandate to establish a level andcompetitive playing field. The right to interconnectwas included in various national laws, and national reg-ulators were authorized to effectively enforce itsimplementation. As expected, regulating interconnec-tion has been one of their main activities.To sum up,the key cross-cutting questions are: (a) Which firms aresubject to interconnection rules? (b) What networkcomponents are subject to the rules? (c) On whatterms can these specific components be shared withcompetitors? (d) What is the most appropriate term forinterconnection contracts? (e) How should the net-work owners be compensated for interconnection to,and re-use of, their embedded systems?

Implementing interconnection regulations raisesmultifaceted issues. First, there are technical-relatedconsiderations that supersede the effective implementa-tion of interconnection agreements. Are interconnec-

tion demands presented by new entrants reasonable? Ifso, can they be met by the incumbent in a reasonabletime frame? Second, the interconnection agreementsare assumed to be freely negotiated, as with any othercommercial contracts, although they cannot becomeeffective unless approved by the regulator. Are partieslikely to reach a fair agreement? If so, how much timeshould be given to them to reach such agreements? Isthe regulator well equipped to review and ratify agree-ments resulting from private negotiations? Whenevernegotiations fail, is the regulator equipped to effectivelyarbitrate ensuing disputes? What should the procedurebe, and what should the appeal procedure be? There arelaws stipulating that interconnection tariffs must be costoriented to ensure productive efficiency. In practice,newly created regulators are ill equipped to performthat role. Field experience also shows that most inter-connection disputes are related to rates,highlighting theoverall importance of the economic dimension ofinterconnection pricing.The remainder of this guide-book treats that dimension.

Essential FacilitiesNormally, at an early stage of competition develop-ment, the incumbent will enjoy a “Stackelberg firstmove advantage,” and will accordingly be consideredto be a dominant player. In such a situation,because theincumbent controls access to inputs that new entrantsneed in order to compete, and can prevent or hampercompetition development in downstream markets,there is a need for specific regulations. As a result, theaccess to these facilities must be regulated.The essential

3 Cost Modeling Principles

facilities concept refers to those facilities owned orcontrolled by one of the players, and whose duplicationis too costly and uneconomical for new entrants.Essential facilities should therefore be mutual or sharedresources, and the need to ensure equal access isstraightforward in certain infrastructure sectors as weillustrate here.

Let us assume that there are two airlines in countryA.Airline X was the only provider of public air trans-port service before the government decided to liberal-ize the market. In the aftermath of air transportliberalization, the government licensed a new airline Y.Prior to the liberalization, airline X was operating 50routes from the main airport built and owned by thecompany. To compete with X, airline Y needs todevelop its airport or negotiate access to airline X’s air-port.What would be the outcome of the competition?It is likely that airline Y’s capacity to compete with Xwill be seriously undermined unless it gains fair accessto the airport facility controlled by X. In conclusion,the new legislation that enacted the liberalization ofthe airline market must also ensure fair access to theairport facilities controlled by X. Then, the questionbecomes how to set the terms for providing and regu-lating this access.

Obviously, operations and airport ownershipshould be unbundled to allow fair competitionbetween airlines. Consequently, airport infrastructuresare usually owned by the state or its representativewhile the management of the facility is often delegatedto a commercial entity. Unbundling ownership andoperation of the airport facility hence allows for theestablishment of a level playing field for airline serviceprovision. Of course, further regulations are needed toensure that all licensed airlines have equal access to airtransport services. In the absence of such a regulatoryframework, each new airline will have to run a privateairport, or be extremely dependent on facilities ownedand managed by other competing airlines, hence limit-ing the development of competition.

Similarly to the airline industry example discussedabove, some components of the fixed telecommunica-tion network owned and operated by the incumbentcan be considered as being essential facilities, at leastduring the early development stage of competition.Anyfailure to provide new entrants access to these facilities

drastically limits the scope of competition. Indeed, inthe absence of an interconnection regulatory frame-work, new entrants would have no other alternativethan to roll out their own local loop network in orderto reach existing telephone consumers. Consequently,consumers would need to subscribe to as many serviceproviders as possible, and the overall social value of thetelephone system would be suboptimal.

Pricing of Essential FacilitiesAlthough interconnection is a mandatory obligationembodied in national regulations, concluding andmanaging interconnection agreements are generallyleft to concerned parties, as are other commercialagreements.1 Consequently, the obligations mandatedby legislation and regulations are considerably weak-ened when the parties’ objectives differ significantly. Itis, therefore, important that regulations provide soundsafeguard measures protecting the most vulnerableplayers from anticompetitive practices by the incum-bent. During the early stage of competition, theincumbent’s survival does not depend on fair access tocompetitors’ networks. The survival of new competi-tors, however, is highly dependent on the terms ofaccess to the incumbent’s network and customers.During, the interconnection agreement negotiations,incumbents will likely charge high prices for access totheir networks, while entrants will seek cheaper prices.Incumbents will also try to delay the provision ofinterconnection services, as long as possible, to com-petitors. In other words, the regulator’s role is crucial indeciding fair access terms.

Specific regulations are needed, and are justified,whenever one of the market players enjoys a dominantposition. However, in an extremely rare situation inwhich there is no dominant player, and assuming thatall new entrants have sound and mutual interests inaccessing each other’s network at competitive terms,enforcing specific interconnection regulations may beneedless. Nonetheless, the most common situation iswhen one of the players dominates some market seg-ments, but not necessarily all of them. In that case, thedominant player finds no incentive in facilitating theconclusion of fair interconnection agreements.

It is essential to spell out dominance criteria, asclearly as possible, and to outline the obligations of the


12


13

dominant player.The new European regulatory pack-age (2002) specifies that “an undertaking shall bedeemed to have significant power if, either individuallyor jointly with others, it enjoys a position of economicstrength affording it the power to behave, to an appre-ciable extent, independently of competitors, customersand ultimately consumers. Where an undertaking hassignificant market power on a specific market, it mayalso be deemed to have significant market power on aclosely related market,where the links between the twomarkets are such as allow the market power held in onemarket to be leveraged into the other market, therebystrengthening the market power of the undertaking.”2

In practice, controlling a market share of greaterthan 25 percent is often considered as being in a domi-nance situation.3 In general, it is recommended to referto any relevant jurisprudence that national competitioncommissions or authorities would have established inthat area.As a result, fixed telecommunications incum-bent operators would be rightly considered as beingpowerful or dominant operators.Whenever an operatoris dominant or powerful, it is more likely to implementtariff discrimination, hence distorting competition. It istherefore essential to ensure that interconnection ser-vices are provided on a nondiscriminatory basis. Inter-connection services provided to new entrants should beidentical, in terms of quality, technical conditions, andrates to similar services that an incumbent would beproviding to its own affiliates.4

Regulating interconnection agreements impliesthat the regulator be appropriately staffed andequipped with a variety of regulatory tools that makeit possible to:• Effectively enforce the accounting separation prin-

ciple or the unbundling of regulated activities fromthe competitive ones.

• Ensure that interconnection rates are nondiscrimi-natory.5

• Publish a detailed reference interconnection offer(RIO), including the description of relevant offersbroken down into network elements as demandedby the market, and complemented by the corre-sponding modalities, conditions, and prices.

• Ratify interconnection reference offers submittedby dominant operators according detailed proce-dures.

• Ratify the terms and conditions of negotiatedinterconnection agreements.

• Effectively arbitrate interconnection disputes.

Cost OrientationNondiscrimination and transparency are not the onlyobligations imposed on dominant operators. Intercon-nection rates are also required to be cost oriented.There are two ways for competition to blossom in thetelecommunications industry. One way is to supportinfrastructure-based competition. According to thatway, policies and regulations are developed with incen-tives to promote investments in infrastructure. Therationale in this strategy is that without affordableinfrastructure, competition in service provision willremain limited. It is therefore important to support thedevelopment of alternative infrastructure providers;this will make the best possible use of technologicalinnovations and will lower entry costs to serviceproviders.The second way is to support service-basedcompetition.According to that way, policies and regu-lations should be developed with incentives to pro-mote investments in service provision that maximizethe load factor of existing infrastructure.

Regulation of input prices is justified wheneverthere is a risk of substantial anticompetitive practices.Input prices can be priced at excessively high levels orcan be priced at overly low levels. In both cases, com-petition is limited or constrained, either downstream orupstream. However,“the imposition of a price controlby national regulatory authorities should not have anegative effect on long-term competition,nor discour-age investment in different infrastructures.The nationalregulatory authorities should take into account theinvestments made by operators providing these inputs,factoring in the risks incurred accordingly.”6

In practice, the cost orientation principle does not meanselling at marginal cost. It is about determining the average costincurred by an efficient operator using the best available tech-nology. Consequently, the above-mentioned averagecost incorporates possible economies of scale and scopeachieved by the operator providing the input.The over-all objective is, therefore, to ensure that input pricesreflect the industry productive efficiency frontier.

To ensure the orientation toward costs, there aretwo main approaches:

• The first approach relies on benchmarking inputprices in similar or comparable environments.Thisapproach contains a serious shortfall. In general,economic conditions differ from one country toanother, and differences cannot always be explainedby market factors. Some of the differences couldjust be related to the geography or other specificsocioeconomic conditions. Consequently, thisapproach only provides rough estimates of possiblecost frontiers. Regulators should refrain from rely-ing excessively on benchmarking to set intercon-nection rates.

• The alternative approach is analytical and reviewsthe cost structure of the regulated operator againstthe one provided by an efficient operator. Becausethe operator provides a broad range of services, it isimportant to be able to differentiate costs accordingto their relevance to the specific input considered.To ensure the cost orientation of interconnection

rates submitted by dominant operators, the regulatormust build and enhance its knowledge with respect to theindustry cost frontier and cost drivers.The next section fur-ther develops the cost allocation methodologies andhighlights their relevance and limitations.

Cost Determination MethodologiesFor multiproduct firms, determining the cost incurredto produce a specific product or service is a delicate andcomplex exercise. With the limitations that apply, thissection presents the different methodologies that couldbe used to assess a multiproduct firm cost structure.

The Generic Costs Borne by a Multiproduct FirmFigure 3.1 illustrates the complexity of a multiproductfirm cost structure. It is assumed that the firm manu-factures five products (A, B, C, D, and E).The overalltotal cost incurred by the firm is also known.The issue,therefore, is to determine the total cost incurred inproducing each product. How should the unit cost perproduct be determined? To begin with, it is importantto define the following concepts:

Direct costs or directly attributable costs areexpenses that are incurred when producing a specificservice or a series of services or products. In otherterms, direct costs attributed to product A will cease toexist if product A is no longer manufactured or pro-

duced by the firm. Consequently, these expenses aretied to the production of a specific service or productand should not exist if that production is stopped.Direct costs can be fixed or variable.

Joint costs are generated by a family of services orproducts (for example, buildings costs for a telephonefirm). From an economic viewpoint, joint costs arecosts incurred in fixed proportions7 every time a ser-vice or a product belonging to the same family is pro-duced by the firm. For example, a telephone companyincurs joint costs whenever it conveys a local, interur-ban, or international call.

Common costs are shared by all the services orproducts of the company (for example, the fixed costsof acquiring licenses). Common costs include theremainder of the costs that are not directly attributableor joint, and which are incurred by the firm.

In conclusion, summing up joint and commoncosts boils down to the total shared costs incurred bythe firm. These costs can be attributed to services orproducts manufactured by the firm using more or lessarbitrary criteria. However, whenever shared costs canbe attributed in a nonarbitrary way, reflecting thecausality factor, they are referred to as indirectly attribut-able costs. Conversely, whenever the attribution canonly be arbitrary, it is referred to as nonattributable costs.


14

Joint costs Joint costs

Common costs

Variable costs Variable costs

Fixed costs Fixed costs

Product Product Product Product ProductA B C D E

Directly attrributable costs

Figure 3.1 The Costs of a Multiproduct Firm:An Example for Five Products

Within directly attributable costs, it is important to dif-ferentiate fixed from variable costs as follows:

Fixed costs represent the proportion of the firm’sexpenses that does not depend on, or vary with, theactivity of the firm. Fixed costs include productioncapacity costs and other preinvestment expensesincurred when preparing the launch of the firm’sactivities. In the event there is a major variation of thefirm’s activities, the fixed costs component will alsovary as a result of the capacity adjustment. However,these adjustments are not necessarily below specifiedthresholds. From an economic viewpoint, fixed costsare assumed to be independent of the volume of pro-duction and are borne by the firm even if it is notoperating. Whenever the firm’s activities are shutdown, some of the fixed costs incurred by the firmbecome sunk costs. Sunk costs are considered to benonrecoverable after the firm’s activities cease.

Variable costs are closely related to the level andthe development of the firm’s production and market-ing operations.When some operations are halted, thecorresponding variable expenses disappear. Con-versely, when operations develop, variable costs alsomove in the same direction.Variable costs include rawmaterial costs, labor costs, other intermediary inputcosts, as well as variable marketing costs (deliveryexpenses, brokerage, commissions, allowances). Vari-able costs are not strictly proportional to the develop-ment of the activity because of the evolutioncharacterizing production factors or technology inno-vation. For instance, if the raw material costs vary pro-portionally to production, that is not the case withsalary costs.

The sum of the fixed costs, the variable costs,the joint costs, and the common costs gives thetotal production cost or global cost. The globalcost or total production cost is directly related to theproduction volume (total cost increases with produc-tion increase). However, in the presence of scaleeconomies, the unit cost drops as production increases.Whenever scope economies are present, it is econom-ically more efficient to have only one firm serving themarket than to have several competing firms.

Two fundamental cost concepts yield from the totalcost definition recalled above: the average cost and themarginal cost.

Average cost is the unit cost obtained by dividingthe total cost by the number of units produced. Theaverage cost function will decrease as productionincreases up to a threshold beyond which it thenincreases (at least in the short term) with the output.The average total cost corresponds to the sum of thevariable average costs and the average fixed costs.

Marginal cost is the total cost variation resultingfrom a variation of the firm’s production.Marginal costis defined by economists as the incremental cost result-ing from the production of an additional unit (or costof the last unit produced).A more formal definition isgiven by the first derivative of the total cost functionrelative to the produced quantity.

Average and marginal costs are basic concepts ineconomics, and the definitions given above hold formonoproduct firms. In summary, it is important torecall that marginal cost represents the theoretical bot-tom cost that a firm has to recover in the short run.

These definitions are slightly modified for a multi-product firm. In fact, the total costs for a firm produc-ing several goods depend on the quantities and theproportions of goods produced. It is therefore impor-tant to differentiate two polar situations: (a) propor-tions of produced goods do not change; (b)proportions of produced goods do change. The con-cepts of radial and incremental cost are then used torefine marginal and average costs.

Average radial cost. Whenever the family ofgoods produced by the firm remains unchanged dur-ing the production cycle, it is more appropriate to usethe concept of the average radial cost instead of averagecost (that is, with a constant proportion of products).

Average incremental cost. Whenever there is achange in the composition or proportion within thefamily of goods produced by a firm, it is appropriate touse the concept of average incremental cost.The aver-age incremental cost is defined as the average cost asso-ciated with a product or a group of products amongthose manufactured by the firm. The average incre-mental cost for a product group usually decreases withthe increase in the number of product groups (scopeeconomies).

In theory, the marginal, radial, and incremental costsrefer solely to the variable component of the cost func-tion. Pricing at marginal cost does not enable the firm


15

to recover the fixed cost. This situation occurs in anindustry in which scale and scope economies prevail.One way to solve this problem is to use long-termincremental average costs, as all costs are then variables.However, joint and common costs still have to befinanced. Other important cost concepts exist, but wedo not discuss them here. Instead,we refer the reader tomore advanced economics textbooks.8

The cost typology, discussed above,has been simpli-fied on purpose to illustrate issues and problems per-taining to cost-allocation methodologies when dealingwith a multiproduct firm. Conversely, the presentationdoes not reflect the refined cost allocation criteriafound in most recent economic literature.

Economic Criteria for Assessing CostsTaking into account the economic dimension of thevarious costs categories, as discussed above, entails abetter understanding and definition of cost allocationcriteria.

The difference between fixed and variable costs istime related. Generally, fixed costs are long term, in thesense that they reflect expenses incurred by the firm todevelop capacity and meet its production objectives.Conversely, variable costs are directly related to theday-to-day operations of the firm;hence, they are shortterm. However, in the long run, even fixed costs arevariables. It is therefore possible to reconcile costaccounting analysis to the economic analysis.

NOTIONS OF ECONOMIC COST. The notion of eco-nomic cost involves bringing a series of costs spreadout over time back to one base year.Adding up all ofthese costs does not make it possible to measure theireconomic importance.As such, if we have an incomeamount Q in year 0, it would be expected that if thisamount were invested according to market conditions(interest rate i), it would yield an income Q∗ suchthat Q∗ = Q × (1 + i) in year 1, and Q∗ = Q × (1 +i)2 in year 2, and Q∗ = Q × (1 + i)n in year n, etc.

Conversely, if one intends to spend an incomeamount D in year n, that implies he or she should,today, have an income equivalent to D/(1 + i)n today.A sound measurement of a series of expenses with dif-ferent time occurrence requires discounting theexpenses (that is, dividing the expenses in year n by the

(1 + i)n before adding them. Above, i represents thecost of capital, that is, the cost applying to resourcesborrowed from the financial market or resources pro-vided by the shareholders in terms of equity.

We also refer to the following:• Total discounted cost, which is given by [the

initial investment9] – [the resale value in the dis-counted terminal year] + [discounted operatingcosts].

• Economic cost or average discounted cost,which is the constant annuity equivalent to the dis-counted current cost.

• Discounted marginal cost for year n, in theabsence of a resale value, is equal to the operatingcosts. If the resale value is not 0, it is the sum of theoperating costs and gap between the discountedresale value in year n – 1 and resale value in year n.

TRENDS ON COST DYNAMIC EVOLUTION. Historicalcost is the cost value entered in the firm’s books (pur-chasing cost or production cost). This cost obviouslycannot represent the real cost of the asset at the end ofseveral years for a series of reasons, including wear,obsolescence, depreciation of the currency, and aging.

Forward-looking, long-term cost: Contrary tohistorical cost, forward-looking cost relies on the bestavailable technologies, assuming that the firm’s produc-tion cycle is globally at optimum.

As a general rule, for a company producing severaloutputs, provided that it is less expensive to jointly pro-duce all the outputs than to produce them separately,the total cost for producing one output is lower in ajoint production structure than in a stand-alone pro-duction. Otherwise, it would be in the company’sinterest to produce these different outputs separately.Joint and common costs, therefore, illustrate economiesof scope, which the industry production structureentails. However, allocating these costs among productsremains a difficult and complex task.

There are several methodologies in regard to allo-cation of joint and common costs. However, none ofthese methodologies provides a satisfactory solution.Depending on the methodology used, the resultingallocation is biased by the arbitrariness embodied ineach criterion. In general, we look for a cost distribu-tion that avoids cross-subsidization among services or


16

products to the greatest extent possible.As discussed inthe box below, the concept of cross-subsidization alsohas several definitions, depending on the context.

There are clear-cut differences in the way the allo-cation of joint and common costs is handled by variousmethodologies. Four main cost allocation methodolo-gies are available. Some are cost allocation methods inthe strict sense of the term, while others are pricingmethods from which allocation principles are derived.The four methodologies are:1. The fully distributed costs (FDC) methodology.2. The efficient component pricing rule (ECPR)

methodology.3. The Ramsey-Boiteux and Laffont-Tirole method-

ology, which rely on demand-price elasticity.4. Long-term incremental costs (LRIC) methodol-

ogy.None of the above methodologies is fully satisfac-

tory. All of them build on sound economic rationale,and can be criticized or defended depending on one’spoint of view.The cost model proposed herein buildson the long-term incremental costs methodology,which will be developed later.

The FDC methodology records the expensesincurred by a firm and allocates them to respectiveproducts based on the causality principle. In thatmethodology, a cost breakdown procedure is used thatgroups costs by nature, function, and, according to anintertwined nomenclature, hierarchy. This approachrelies heavily on the availability of reliable accountinginformation, which is usually generated by activity-

based accounting (ABC) systems. Furthermore, theallocation of joint and common costs is done usingarbitrary distribution keys.

There are two families of LRIC models: (a) top-down LRIC and (b) bottom-up LRIC.

Top-down LRIC is essentially an ABC methodol-ogy.Top-down LRIC models derive incremental costsby summing up the costs that can be directly attributedto the service and adding a markup that covers a pro-portion of joint and common costs. The markup isdetermined with the help of arbitrary distributionkeys. Therefore, the resulting cost is more backwardoriented than forward looking.

Bottom-up LRIC is a constructivist methodologyfor determining forward-looking service cost. Themethodology involves simulating the cost incurred byan efficiently operated network, using best availabletechnology, to provide the service. Bottom-up LRICmodels are highly recommended for regulatory deci-sions, though highly criticized for their lack of realism.The LRIC methodology takes into account fixed costscaused by the provision of interconnection services,but does not take into account the common costs,which do not vary proportionately with the provisionof interconnection services. In practice, the implemen-tation of LRIC models requires a systematic assessmentof demand and cost evolutions, as well as the interde-pendencies that may result.A systematic analysis of net-work element load factors is also critical to accuratelysize the resources needed to convey the incrementaltraffic. In other words, the LRIC methodology makes


17

Box 3.1 Different Definitions Selected for Cross-Subsidization10

• The public policy view: From a public policy perspective, cross-subsidization occurs in a regulated industry when the regulated firmuses revenues from one market to keep operations in another market that is financially not viable.The cross-subsidy is consideredanticompetitive if there are cash flows from noncompetitive to competitive markets.The cross-subsidy is considered a universal ser-vice obligation (USO) if the cash flow (1) goes anticlockwise; (2) occurs only because regulatory rules create it; and (3) would notoccur if the government policy were absent, and/if the markets were competitive.

• The cost allocation view: In more general usage, if a service’s prices do not make a reasonable contribution to overhead costs, itcould be argued that the service is not carrying a fair share of the overheads, and is, therefore, being subsidized.

• The Baumol-Faulhaber view: Baumol and Faulhaber have taken the view that cross-subsidization occurs when prices for a ser-vice do not cover the service’s incremental cost and the company still earns a normal profit (that is, zero economic profit) overall.This implies a maximum price of stand-alone cost.

• A more comprehensive economic view:More recent economic studies have shown that cross-subsidization occurs when pricesfor a service are higher than would be charged by the next most efficient competitor, and the company still earns a normal profit.

it possible to base the cost determination on forward-looking costs and not on historical costs.The next sec-tion provides a more extensive description of theLRIC methodology.

The LRIC ConceptThe LRIC methodology, also sometimes referred to asthe long run average incremental costs (LRAIC)methodology, estimates additional costs incurred inproducing a service, relative to the costs alreadyincurred by producing a portfolio of other services.The incremental costs of a service or element A some-how represent the cost savings, which result from notproducing or not implementing A. In other words, thecosts incurred to produce A over and above the portfo-lio of existing products are considered as the incre-mental cost.

The long-run concept involves taking the costsincurred in a long-term perspective. In the long run,production fixed costs can be considered as being vari-able costs.The long-run incremental costs of a serviceor element A therefore represent all of the costs thatcould be avoided if A were not produced or imple-mented. Hence, the incremental costs include all thecosts directly attributable to A, whether these are vari-able in stricto sensu (depending on the level of traffic ata given capacity) or fixed (making up the capacity).

However,A can also make use of elements, services,or functions needed jointly with other services or ele-ments. The incremental costs (even long run,strictly speaking) only take into account a portionof these costs in the case of joint costs (pro ratatheir incidence), and not common costs.11 Incre-mental costs are important in the sense that they reflectthe company’s decision for producing A. Usually, indeciding to produce A, the firm expects that revenuesgenerated by A should exceed the incremental costincurred and provide an earning income to capitalabove the cost of capital.

Nonetheless, incremental costs, as we have juststrictly defined, are difficult to use in pricing access toa service or network element, to the extent that theyonly cover part of the costs. Given that A also usesother network’s resources and is partly responsible forjoint costs incurred by the firm, it is necessary to allo-cate to A a portion of these joint costs. By adding all

these costs to the generic LRIC, we obtain a total ser-vice (TS) or total element (TE) LRIC. TSLRIC orTELRIC involves allocating pertinent12 joint andcommon costs to A.

Finally, it is worth outlining how the cost allocationis implemented in practice.Two options are generallyconsidered, although either of them can also lead to analternative:1. Historical costs form a first option. The method

therefore involves evaluating the costs on the basisof their accounting values,13 possibly adjusted totake inflation into account.

2. The forward-looking or current costs are the coststhat would be incurred if the production systemwere rebuilt on the date of calculation.

The first approach is often described as a top-downapproach and the second as a bottom-up approach.This is even more essential if we introduce a consider-ation of technical progress in cost evaluation. In fact,the production of A on the date t by an incumbent maynot require specific new investment, but will resultfrom investments already made in previous productioncycles. As a result, working with outdated technology(the network architecture unchanged) or with a betterarchitecture, as available on date t, are options that canbe considered. In retaining the historical productionsystem, we end up using historical costs or currentcosts to value the incremental cost of producing A.Conversely, in retaining the most efficient current pro-duction system, we end up valuing the incrementalcost of producing A with current costs. For practicalreasons, an “average historical” architecture14 is gener-ally selected.This point will be discussed in more detaillater.

We then assume that:1. The operator is an efficient operator that minimizes

costs for a given production volume.2. The costs are current costs.

As we have seen, this accounting and historicalmethod involves breaking down the company’s costsand allocating them among its different products.Whileassigning directly attributable costs to respective prod-ucts is a straightforward issue, allocating joint and com-mon costs is complex and sensitive.The difference between


18

the FDC and the TSLRIC method, using historic accountingcosts and technologies, therefore resides in the nonincorporationof nonrelevant common costs.Beginning with a total costestimate obtained with an FDC approach,we can moveto LRIC, step by step, by removing layers of cost ineffi-ciencies as shown in figure 3.2.

Effective regulation of interconnection costs, there-fore, introduces the following concepts: the specificnature of costs and the pertinence of costs.

The increment generally relates to a group of ser-vices using the same production infrastructure. Thecosts incurred from the provision of a specific serviceof the family should be determined within that frame-work. For example, a telephone company suppliesretail and wholesale services. The cost incurred fromproviding retail services should be derived from theincremental costs incurred from providing wholesaleand retail telephony services. However, the costs formarketing retail telephone services are costs that apply

to retail services and are not shared with the intercon-nection services. Inversely, interconnection servicesgenerate specific costs, which should not be includedin the increment costs (for example, co-location costsfor new entrant’s equipment, costs for links betweenthe incumbent’s network and new entrants’ networks,cost of modifying the information technology [IT] sys-tems, specific billing cost, interconnection manage-ment service cost, and so forth).These costs should bepassed on to interconnection customers only.To enablethe provider to recover the costs incurred, a two-parttariff scheme is usually applied.

The concept of pertinence affects the handling ofjoint and common costs. Joint and common costs canonly be applied to the increment on the condition thatthey are linked to it, either directly or indirectly.Revealing a causal link takes for granted that an in-depth technical review of the cost structure is per-formed. Common costs include research and


19

Fullyallocatedhistorical

costsFDC

Non-pertinentcommoncosts

adjustment

Economiclifetime

adjustment

Efficientsupplier

adjustment

TELRICtop-downhistorical

Currentcosts

adjustment

TELRICFL

bottom-up

Pertinentjoint andcommoncosts

adjustment

LRICFL

LRIC Adjustment Forward Looking Total Increment

Figure 3.2 Transition from Historical Accounting Costs to Economic Costs (LRIC)

Note: FL, forward looking.

development costs, costs associated with headquartersexpenses and the operator’s operational structure, costsfor staff members who are no longer in their positionsor are on leave, costs related to developing the brandname reputation or marketing, and costs associatedwith unused buildings.Among these costs, the regula-tor has to assess and decide which ones are pertinent tobe accounted for.Only pertinent common costs can beallocated proportionally to interconnection services.Figure 3.3 illustrates the distinction between these dif-ferent categories of costs.

In general, there is a risk in overestimating commonfixed costs. For instance, the former monopoly couldargue high common costs to “squeeze” out its com-petitors in downstream retail markets by reporting tointerconnection some of the costs derived from com-petitive activities. In consequence, it is necessary toensure that the common costs attributed to the incre-ment are sound. In any case, the costs attributed shouldreflect the costs borne by the most efficient operators.

In practice, a certain number of questions are raisedin regard to the theoretical framework discussed.Thesequestions are synthesized in the two points below:• Depending on the network services or elements

considered, the TELRIC method can prove to beless favorable to new entrants than the FDCmethod. If a network section A is subject to sub-stantial depreciation, and if current costs for recon-structing A are quite similar, or even higher than thehistorical costs, a TELRIC can lead to a higher cost

than the one that could be derived with an FDCassessment!15

• Depending on assumptions made, the TELRICmethod can lead to a relatively wide range of esti-mates. It is important to recall that directly attribut-able costs result from the service or elementssegmentation assumed in the model.Therefore, themagnitude of joint and common pertinent costsdepends on the segmentation refinement. Further-more, allocating joint and common costs remains amajor hurdle, even though the use of more accuratedistribution criteria can induce smaller cross-subsi-dization between services or products.Selecting specific TELRIC methods leads to some

kind of arbitrage exercise, which is similar to position-ing of a cost cursor within bounded values. The costvalue (obviously high), which is supposed to be favor-able to incumbents,will be at one extreme; at the otherextreme will be the cost value (obviously low) thatfavors new entrants. Two other considerations areadvanced in regard to positioning this cost cursor:• “Theoretical” considerations limit the admissible

spread of cost values, by suggesting that the spreadbe constrained by the implementation of a systemeliminating cross-subsidization, and that resultingestimates be below the separate cost of providingthe same services.

• “Political” considerations related to the appropriateeconomic signal are provided to market players andenable them to arbitrate between investing and pur-chasing the inputs from the incumbent (for exam-ple, the play or pay principle). Two concerns are,however, raised.The first concern is to ensure thatinput prices are not below the costs incurred by theincumbent. The second concern is to ensure thatinput prices allow for efficient allocation ofresources.An excessively low rate for input can lead to cross-

subsidization toward this service and send undesirablesignals to the market.An interconnection price belowcost (in case of an efficient operator) could also endan-ger alternative infrastructure development, as thatshould make new investments less attractive and prof-itable. Conversely, an interconnection price above cost(in case of an efficient operator) is also likely to bias themarket by switching demand to infrastructures that are


20

Specific costs

Increment

Nonspecificcosts

Variable costs

Fixed costs

Joint costsPertinent joint costs

Nonpertinent joint costs

Pertinent common costs

Nonpertinent common costs

+

+

Figure 3.3 Cost Structure

actually less efficient, and does not enhance productiveefficiency in the sector. More specifically, the incum-bent does not have the incentive to shift to a more effi-cient production process.

It is, therefore, up to the regulator, using cost mod-els to determine the “correct” price level for this input.In so doing, the regulator collects and handles variousinformation and has to fix a certain number of themodel’s parameters.These parameters and assumptionsare improved on, with use and time.

Unfortunately, regulators and operators do not havea good knowledge of the industry cost structure.Therefore, the TELRIC bottom-up method representsa decision-making tool that induces improvements onprocedures and processes implemented by regulators oroperators to collect and retrieve cost information.Thisalso improves the quality of investment decisions forfirms, as it naturally enhances the transparency of theinterconnection services market.

The following section reviews questions raised bythe LRIC method for determining interconnectioncost.The model deals with the issue of economic pric-ing of interconnection services and does not coverissues related to co-location or management servicesthat are associated with interconnection.

Advantages and Drawbacks of LRICThe following subsection summarizes the main advan-tages and drawbacks of the LRIC methodology.Withthese comments in mind, we expect the user to be ableto outline the limitations of the LRIC methodologyand notice how its implementation can help in sortingout and, probably, in clarifying a complex and sensitivesubject.

One of the most effective features attached toLRIC pricing schemes is the sharing of the productiv-ity gains that the various market players could derive.In so doing, implementing LRIC pricing schemesimpedes excessive profits by the interconnection ser-vice provider. The relevance of the LRIC methodol-ogy therefore depends on the efficiency concept. In sodoing, interconnection rates are derived from a bench-mark provided by an efficient operator. Cost modelsare, indeed, developed to simulate, with some accuracy,the cost frontier that could prevail in a specific eco-nomic and market environment. Interconnection rates

must, therefore, be equal to long-run incremental costsin order to maximize economic efficiency. LRIC pric-ing schemes are forward looking and provide betterincentive for static cost efficiency. In a dynamic frame-work, the impact of LRIC methodology in determin-ing interconnection rates remains inconclusive (Laffontand Tirole 2001).

Tight access pricing regulation prevents the incum-bent, controlling essential facilities, from extracting allthe monopoly rents related to its dominant positionover these resources. Although these regulationsencourage the efficient utilization of the resources,they can also discourage further investments, subse-quently causing significant social welfare losses.This isthe most probable risk that could be implied by inef-fective access pricing regulation.

The LRIC methodology has been criticized byseveral authors. Salinger (1998) and Laffont and Tirole(2001) argued that LRIC regulation provides the reg-ulators with a key tool to manage industry entry. It is,therefore, crucial to ensure that, whenever a mistake ismade, it is made in favor of overinvestment rather thanunderinvestment.

In a long-run framework, firms can reassign theirinputs according to input price and output.Wheneveran output is unprofitable, firms can freely dispose cor-responding assets by selling them or by dispatchingthem to other activities. However, in the telecommu-nications industry, most assets cannot be freely dis-posed, because exiting an activity is costly.As shown byHausmann (1996), when the regulatory frameworkallows entrants to exit and divest some of their assets,and does not provide similar options to the incumbent,entrants are less likely to invest in building their owninfrastructure.

Furthermore, the value of most telecommunica-tions operators’ assets with current available technolo-gies is lower than their corresponding book value.Consequently, entrants have no incentive to build theirown networks and will prefer using the incumbent’sinfrastructure. LRIC can, thus, introduce inappropriateincentives for entry, because of probable cross-subsi-dization from the incumbents to entrants.16

It is also worth pointing out that the determinationof long-run incremental costs remains after all discre-tionary. Salinger (1998) suggests that the use of LRIC


21


22

is theoretically sound, but its implementation in prac-tice is rather complex and could undermine the prof-itability of the incumbent’s investment if poorlyexecuted. Similarly, Valleti (2001) argues that accesscharges based purely on LRIC are an appropriatebenchmark when retail-level distortions are eliminatedor dealt with effectively by other regulatory instru-ments. Consequently, when the incumbent’s retailprices are not cost oriented because of universal serviceobligations or delays in rebalancing retail rates, there isa need to add a uniform markup to the LRIC esti-mates, although doing that does not reflect any soundeconomic analysis.

Another problem is related to the pertinence of theessential facilities concept in the telecommunicationsindustry. Celani, Petrecolla, and Ruzzier (2002)pointed out the following problems that are not prop-erly addressed.What telecommunications market seg-ments could qualify as essential facilities? What assetscan be considered as being essential facilities? This callsfor a sound methodology that enables disaggregationof the access market segment in order to identify thoseeligible for the essential facility concept. Obligationscould then be limited to these market segments.

In developing countries, the key market players aremostly multinationals. Regulators should thereforeclosely monitor how transfer pricing schemes areimplemented between affiliates and mother compa-nies. Similarly, an appropriate database on cost infor-mation should be developed to limit the scope ofopportunistic behaviors.

Definition of Interconnection ServicesInterconnection services are offered to operators(fixed-line or mobile network operators) in order tocollect or terminate their traffic from or to other com-peting operators. In general, four major categories oftraffic are distinguished, as illustrated below:

Traditionally, three possibilities are considered inregard to the network’s nodes that handle the trafficcollection or termination:• When the interconnection is handled at a local

switch, the service is termed local.• When the interconnection is handled at a transit

switch, with the latter serving only a limited orspecified transit area, the service is termed singletransit.

• When the interconnection is handled at a transitswitch,with the latter serving or providing access tothe overall network’s nodes, the service is termeddouble transit.In principle, whenever a network’s topology and

architecture are compared, the collection and termina-tion costs are identical.The equality is not sound, but isestablished in principle. In fact, the lack of reliable dataon possible routing alternatives does not allow for fur-ther distinction of the cost of handling a terminating ora collected call.

In other words, five interconnection services aregenerally considered:

Domestic transit is generally calculated as the dif-ference between double transit and single transit. Inter-national transit is a surcharge applicable tointernational calls.

The following diagram shows the entry points,depending on the nature of the interconnection ser-vice. International service as assumed in the model isrouted via an international transit switch providing anentry point in the transit area covered. If the entrypoint is located in a different area, the cost of domestictransit has to be added.

Origin–Destination

Recipient of Interconnection Supplier of InterconnectionTraffic collection Traffic origination/termination

Traffic termination Traffic origination

Origin–Destination

Interface Point Correspondent NetworkLocal switch LocalTransit switch Single

in the area transitTransit switch in a Double Domestic

different area transit transitInternational International

switch transit

Definition of the IncrementA telecommunications operator generally offers a widerange of services, whether to its subscribers or to otheroperators.These services have been traditionally classi-fied according to their commercial nature. Forinstance, one usually differentiates local from long dis-tance and from international services.Added-value ser-vices are also differentiated from plain old telephonyservices. However, another service classificationapproach is based on the use of network elements fur-ther described below. In general, there are three majorcategories of network elements:• Elements that are dedicated to end users (which are

used solely by a subscriber). These elements aregenerally found in the local loop network or accessnetwork and include a subscriber’s connection lineto the switch;

• Elements that are shared among users (dedicated tousers on a dynamic basis upon request).These net-work elements are allocated temporarily to a sub-scriber and are generally found in the coretelephony network.

• Elements that are shared by users but are used forthe provision of complementary or supplementaryservices (telephone card services or other servicesoffered by intelligent network features).The three services discussed above can be labeled

as: (a) access services, for services dedicated to eachuser, (b) transport services, for services shared among anetwork’s users and offered by the core network func-

tions, and (c) value-added services, for shared servicesoffered by intelligent functions of the network.Trans-port services can be broken down in various elemen-tary services, according to traffic collection anddelivery points. A correlation can be establishedbetween these elementary services, and the servicesmarketed by the operator. Such a correlation is illus-trated in the table below:

As shown in the matrix, there is a correlationbetween the operator’s elementary services (repre-sented by horizontal lines in the first column) and ser-vices sold to end users (represented by columns in thefirst line). The matrix provides, for each service, thecost components incurred for its provision. Of course,the matrix is not exhaustive. A column representingleased lines could also be added to it. For instance, theprovision of interconnection services involves trans-port services (local, simple, double transit). Conversely,the provision of a local communication involves inaddition to transport services, the retail sale service.

Furthermore, an operator may wish to implement aretail pricing strategy that is not necessarily cost ori-ented, at least, as an entry strategy to develop its busi-ness line. In such situations, the operator will likely optfor a pricing strategy featuring high connection feesand low rental charges. Sector regulation should pre-serve the pricing freedom as long as it does not intro-duce a squeeze and distort competition (that is, as long


23

Figure 3.4 Entry Points of Interconnection Service

Double transit

Singletransit

International

Domestictransit

Entry point

Exit point

Transit area

Transitarea

Local area

Localarea

Fixed local

International

Subscrip- Longtion Local Distance Supple- Intercon-

Connec- Communi- Communi- mentary nectiontion cation cation Service Service

Retail sales service X X X XAdded-valueservice XTransport service

Local X X XSingle transit X XDouble transit X X

Access service X X

Table 3.1 Correlation Matrix between Retail Services and Elementary Services

as it does not eliminate new entrants from the market).It is therefore important to define and ensure the con-sistency between the pricing of the inputs needed toprovide retail services and the pricing of the retail ser-vices. The table below provides an illustration of thelinkages to help process the consistency review.

The retail sales activity includes all the functionsassociated with providing the service and managingthe customer relationship.The retail sales activity pur-chases all the horizontal services required to providethe final service that the end user consumes. In sodoing, it implicitly transforms respective cost elementsused into pricing components. In checking the coststructure of regulated services, the regulator must pay spe-cific attention to the squeeze that could materialize as a resultof undue cross-subsidization of competitively provided servicesby noncompetitively provided ones. The regulator shouldalso ensure that interconnection services, which arewholesale services, are priced below correspondingretail services.

As discussed earlier, the incremental cost methodol-ogy distinguishes directly attributable from nonattrib-utable costs. It involves ascertaining,whether or not theproduction cost of “horizontal” service is increased,whenever a “vertical” service is added to the basket ofother services produced.

Considering the example of interconnection ser-vices, it is important to underscore the following threeconclusions:• The provision of interconnection services does not

modify the retail sales service, as the latter only con-cerns sales to final subscribers.

• The provision of interconnection services does notmodify the access service, as the capacity imple-

mented and its maintenance do not have to beadjusted to bear the flow of this additional traffic.

• In contrast, transport services are affected by thetraffic resulting from interconnection services.As a result, the cost of access services must be cov-

ered by retail service revenues. Unless these servicesrates are rebalanced, there are risks for anticompetitivepractices and cream skimming. In other words, marketderegulation imposes a certain cost orientation fordominant operators’ tariffs, and, consequently, implies aprice restructuring that eliminates the largest existingpricing averaging.17

The core network is used to provide a broad rangeof services. Apart from interconnection services, itfacilitates the provision of retail services, including theprovision of other services such as leased lines. Leasedlines are either aimed at final customers (companies fortheir corporate networks), or at other telecommunica-tions that operators can use to roll their own networks.The operator can also use leased lines for its own inter-nal consumption (to operate another network such asthe telex network,packet data transmission network,oreven in many countries to carry the radio and televi-sion programs).

It is also important to note that the core networkusually shares resources with other networks managedby the operator, or with the access network whoseducts are frequently shared in urban areas. It is impor-tant to report to the core network portions of the jointcosts incurred by other networks and, in particular, theaccess network.

In conclusion, the increment to be considered forthe calculation of interconnection service costs com-prises the core network and excludes the access net-work’s elements that are dedicated to users.


24

Retail Subscription Local Long Distance Supplementary Interconnection Sales Connection Communication Communication Service Service

Retail sales activity X X X XAdded-value service XTransport service

Local X XSingle transit X XDouble transit X X

Access service X

Table 3.2 Matrix of Inputs Needed to Provide Retail Services

Application to African TelecommunicationsNetworks

African telecommunications networks are specificbecause of their early development stage. ExcludingSouth Africa, we can define a representative Africancountry (Afriland) with respect to its telecommunica-tions common specificities.To do that, we aggregateddata for a sample of 40 countries and were able toderive relevant common features, which are embodiedin this cost model.18

The Average Size of an African TelecommunicationsNetworkAfriland, as an African representative country, is charac-terized as follows:

In terms of size,Afriland is larger though less pop-ulated than an average European country. Afriland is

double the size of an average European country. Itspopulation density is four times less, implying higherrollout and maintenance costs for telecommunicationsinfrastructure than in densely populated countries.Furthermore, its average income per capita is 70 timeslower than in EU countries, and the fixed teledensity is100 times lower. The average consumption of tele-phone services is estimated at slightly more than 20minutes per inhabitant per year. On average, only 20percent to 25 percent of the population has access to atelephone within walking distance.

In terms of infrastructure, the coverage of the fixed-line network remains rather limited and connectsabout 75,000 lines. The number of subscribers con-nected to the mobile networks exceeds the subscriberbase of the fixed network and is approaching 100,000.The mobile segment is served by two or three opera-tors offering essentially prepaid services (95 percent ofthe consumers of mobile are prepaid).

Furthermore, approximately 80 percent of thestock of subscribers and 80 percent of the traffic isconcentrated in the capital city. On the whole, theoutreach of the network is limited to the mostimportant urban centers. Extending the coverage torural communities remains a major challenge for theyears to come, as rural areas are also those with thelowest income and population density.As a matter offact, 80 percent of the stock of telephones serves only20 percent of the population. Consequently, if theteledensity is about 2 percent in the main urban cen-ters of Afriland, it is only about 0.125 percent in therest of the country, though the overall national tele-density is 0.5 percent. Afriland is characterized by avery low level of telephone service demand, which isa direct consequence of the limited income percapita.

The implications in terms of network architecture,topology, and costs are:• The transit network is almost nonexistent. Usually,

the transit functions are implemented in localexchanges or switches and are located principally inthe capital and in two or three main cities.

• The transmission network has an unusual topology.When SDH rings are set up, they are used to con-nect telephone switches and their RCUs within thesame urban center or region.


25

LowIncome,

Low EuropeanTeledensity Union Average

Situation Africa (40 (15 EU in 1999 countries) countries) Afriland Country Ratio

Area 20,927,000 3,191,000 525,000 215,600 km2 km2 km2 km2 0.41

Popula- 600 376 15 25 tion million million million million 1.67Popula- lation Density(inhabi-ants/km2) 29 117 29 117 4GDP/capita 300 a 21,000 a 300 a 21,000 a 70TotalGDP(billions ofeuros) 180 7,900 4.5 525 117Stock ofmainlines 3,140,000 200,000,000 75,000 13,300,000 180Tele-density(fixed) 0.5% 53.2% 0.5% 53.2% 100Density of lines/ km2 0.15 61.8 0.14 61.8 440

Table 3.3 Snapshot of Telecommunications Development in Africa in 1999

Source: BIPE 2001.

• The networks still reflect choices of technologymade in the 1990s. The number of small digitalswitches operated is higher than what could beachieved by using modern technology.With the besttechnology available today, most of these networkscould be equipped with only a couple (one or two)of digital switches.The scope of inefficiencies result-ing from this technology legacy is rather great.

• To reach remote rural areas, subscriber concentra-tion systems (TDMAs) are installed.These systemsserve from 8 to 256 subscribers and use time divi-sion multiplexing technology, known as TDMA.TDMA cost per line is particularly high.19

• Fiber-optic systems are not yet massively rolled out.In many cases, the demand does not justify the roll-out of broadband transmission systems; however,more and more operators are beginning to replacetheir microwave transmission links with fiber-opticcables.

• Finally, in some countries, domestic satellite net-works are operated as a means of extending thecoverage to isolated areas (rain forest, islands,desert).The proposed model seeks to account for all these

specific features so that the cost calculation can be asclose as possible to the reality from which should bederived the industry cost frontier. As an illustration,TDMA systems have been included in the increment,though one could argue against inclusion. We haveassumed,however, that telephony circuits implementedthrough these systems are not dedicated to a user, butare shared among all users of the network, includingthose connected to the RCU.As a result, we considerthat the traffic generated by servicing the interconnec-tion demand (termination, origination) contributes tothe sizing of these systems.

Nevertheless, in costing the interconnection ser-vice, the model provides two options:• A cost “with TDMA systems” included in the

increment.• A cost “without TDMA systems” included in the

increment.The results obtained when including TDMA are

very different from the ones obtained when it is notincluded.We suggest that the regulator decides, basedon national specificity and according to the govern-

ment’s universal access policy, whether or not toinclude the cost related to TDMA systems.

Nodes and LinksA thorough description of the core network is neededbefore its modeling is attempted. This descriptioninvolves determining the types of nodes and the natureof the links between the nodes. In telecommunica-tions, a node is generally characterized by a switchingfunction, while the links form the transmission net-work. In developing the proposed cost model, we haveassumed that an average African network has five typesof nodes and five types of links.

These nodes are:• Subscriber RCUs, which do not have independent

switching command capacity, except in the event ofa breakdown of the connection link with the hostexchange. In the model, the small rural exchangesthat are still in service and electromechanicalexchanges have been considered as RCUs. Therationale is that digital RCUs are more likely toreplace these old technologies.

• LSs to which RCUs are connected. Some of theseexchanges have a transit capacity.

• “Pure” transit switches (TSWs),20 which are rare inthe African context, but are integrated into themodel for the sake of completeness.

• ISs or international transit exchange. Theseexchanges are rarely taken into account in the mod-eling carried out in industrial countries because theinternational transit node is not differentiated fromnational transit21. This is not yet the case in Africawhere the international segment is not yet unbun-dled from the incumbent’s monopoly.

• Finally,TDMA systems lead to two types of nodes:– CSs, which are generally coupled with associated

switches to manage the signaling and the interfacewith the transmission systems component;

– TSs, which are the base stations installed in ruralcommunities to provide the last mile connectionto subscribers.

Similarly, there are five types of transmission links:• RCU-LS links are the transmission links between

RCUs and their host exchange.• LS-LS links are the transmission links between two

different LSs. In some cases, these links could also


26

involve LS links to TSs, to the extent that the tran-sit function is implemented.

• Specific links between transit centers (TS-TS).• Links to ISs (between LS and IS). In certain cases,

the IS function can also be implemented in an LS.• Internal links within TDMA systems, or the links

between the central stations and terminal stations.As a result, we obtain the matrix representation

provided below:

It should be noted that local links (LS-CS, RCU,and the central station in a TDMA system) are notoften taken into account because of the frequent co-location of these nodes.

TransitBecause of the small size of African networks, localexchanges (LSs) generally provide transit function.Whenever the network becomes larger in size, interms of coverage, local areas are generally groupedinto transit areas. The single transit feature involvescollecting or delivering traffic in the transit area.Conversely, double transit involves collecting and ter-minating traffic beyond the initiating transit area.When a network does not have a transit switch, thedistinction between local and single transit becomesirrelevant.

For the purpose of the exercise conducted here, wewill assume the hierarchy between interconnectionservices (intra-LS, single transit, and double transit) isestablished along the incumbent’s local rates zones orderived from the administrative or government terri-torial organization (municipalities, counties, districts,provinces, regions).

The first option links the interconnection serviceshierarchy to the retail pricing structure of the incum-bent,while the second option, although more arbitrary,does not.


27

Nodes IS TSW LS RCU CS TS

ISTSW To IS TSW-TSW TSW-LSLS LS-LS RCU-LS (Local RCU link)CS CS-TSTS

Table 3.4 Matrix of Transmission Links

Notes

1. General interest obligations can be imposed,for example, access to emergency numbers.2.Article 13 of the draft “Directive on a Com-mon Regulatory Framework for ElectronicCommunications Networks and Services.”2001. Com 380, European Commission.3. This measure supposes a prior definitionof the relevant market, that is, the one inwhich power has to be measured. In generalit is a market that brings together productsamong which significant substitutions andcomplementarities are present or possiblefrom the customer point of view.4. Definition from the draft “Directive onAccess to Electronic Communications Net-works and Associated Installations, as well asTheir Interconnection.” 2001. Com 369,European Commission.5.The regulator can impose an obligation ona vertically integrated company to make itsinternal wholesale and transfer prices trans-parent in cases where the market analysis re-veals that the operator concerned suppliesfacilities that are essential to other service

providers, whereas it is itself in competitionwith these in the same market downstream(“Directive on Access to Electronic Com-munications Networks and Associated In-stallations, as well as Their Interconnection.”2001. Com 369, European Commission).6. “Directive on Access to Electronic Com-munications Networks and Associated In-stallations, as well as Their Interconnection.”2001. Com 369, European Commission.7. Strict economic definition.8. For example, notions of opportunity cost.(In our context, a cost that enables the partyselling the resource to a third-party “whole-saler” to obtain remuneration equivalent tothat which it would have obtained by sellingit in the end market.This cost is therefore thesale price less the cost of the retail price. It isequivalent for the supplier to sell on the in-termediary market or on the final market. Insome ways, it is an access cost for resellers.)9. This is assumed to be undertaken here inyear 1 (otherwise, it would be necessary to takethe discounted sequence of investment costs).

10. Mark A. Jamison, available athttp://bear.cba.ufl.edu/centers/purc/primary/jamison/Pricing.pdf.11. Joint costs refer to costs incurred by twoor several products in the same productionprocess, in a constant proportion.We talk ofcommon costs when the costs are incurredby several products and remain unchangedregardless of the relative proportion of theseproducts (the salaries for operators’ head-quarters functions), that is, when a product isoffered, the second product is produced bythe same production without a supplemen-tary cost.12. That is, which demonstrate a causalityrelationship.13. These are also referred to as embeddedcosts.14. By taking over the topology of the his-torical network, that is, the same intercon-nection equipment locations inside the net-work (switching, concentration, distributionand so forth).This is, thus, the scorched nodeoption, which involves retaining the net-


28

work’s real hierarchy and the current trafficmanagement rules. Imagining an optimumnetwork would lead to a certain number ofcriticisms concerning its feasibility, and itspossible operational capacity, on the impactof this virtual architecture on other prices.15. This explains that in several countries,the FDC method was considered as the mostfavorable to new entrants in terms of accessto the local loop, where civil engineeringcosts are dominant, although less favorable interms of interconnection.16. LRIC discriminates between incum-bents and entrants in favor of the latter.

When incumbents make investment deci-sions, technology that the regulator mayconsider to be best may not be available.17. Within this context, for operators whohave not undertaken the required price re-structuring for whatever reason, there maybe a continued need to subsidize some loss-making services from profitable services and,on an almost general basis, a subsidy from in-ternational and long distance calls, to the ac-cess and/or local communication service. Inthis case, two phenomena have to be distin-guished: the interconnection service mustmeasure the value of these cost-oriented ser-

vices and a possible temporary “accessdeficit” (until the price restructuring isachieved) must compensate for the pricingequalization in force.18. This roughly means African countrieswhose gross domestic product (GDP) perinhabitant is lower than EUR 1,000.19. These radio concentrators are presentedin appendix 3.20. Also called a tandem switch.21.This has to do with market liberalization,which has now enabled operators to estab-lish interconnection point of presence acrossthe borders.

29

The model reconstructs a network, as would be doneby an efficient operator using a forward-looking LRICmethodology. The reconstruction of the network isdone in line with the realities of the Sub-SaharanAfrican environment, as discussed earlier. Before dis-cussing the modeling in detail, let us recall the follow-ing principles.

The Modeling Principles

A Long-Run ApproachThe LRIC method adopts a long-run approach. Thereconstruction of all network elements is assumed atyear 1 and includes the increment.

A Forward-Looking ApproachSelecting a forward-looking approach means consider-ing both the best technologies available and their cur-rent costs. From a pragmatic viewpoint, this comesdown to considering the digital technologies availableon the shelves today. Consequently, when modeling,we replace old technology with modern “equivalent”technology, which is more efficient and cost-effective.

For the transmission system, forward lookingimplies selecting SDH systems over fiber-optic cables.Although fiber optic is considered to be the most flex-ible, efficient, and economic technology in regard tothe bandwidth unit cost, it is not yet rolled out system-atically in Africa.

As regards the costs to be taken into account, thesehave to be estimated at the current acquisition price,and not at book value.As such, the decision whether

to invest or purchase services (pay or play) should bemade in the light of the prevailing economic situa-tion.

An Efficient ApproachThe cost modeling must reflect decisions made by anefficient operator, producing the increment services atthe best cost while taking into account the technolo-gies available. As a result, the model simulates a net-work that, at a given production level, minimizes thetotal cost by using best available technologies.

This requirement raises a certain number of ques-tions with respect to the network architecture. As wehave already stated, an incumbent inherits a networktopology, which is largely determined by successivegenerations of technologies.Therefore, in dealing withthe efficiency problem, there are two possibleapproaches:• Modeling a network providing the expected ser-

vices after optimizing its topology and architectureand eliminating the technology legacy.This optionis called “scorched earth.”

• Keeping the existing network’s topology (the loca-tion of network nodes), while reconstructing thenodes and links with the best available technolo-gies.The result is, in a way, a topologically identicalnetwork.This option is called “scorched node.” It isthe one selected in the proposed cost model.If the resulting topology is clearly nonoptimal as it

would be recommended to ensure production effi-ciency, then the regulator can prescribe a better net-work configuration.

4 Modeling Principles

Beyond the choice of the best technologies with anexisting network structure, the question of efficiency asregards network operation also arises.Digital technolo-gies, in general, and, more precisely, current networkmonitoring and management systems, make it possibleto downsize factors of production, buildings space,labor, and so forth. Regulators and operators mustagree on the optimum level of efficiency. LRIC mod-els should not, under any circumstance, take intoaccount excess staffing from previous managementbecause market deregulation should be a strong incen-tive for the incumbent to improve its efficiency.

An Economic Approach, Not an Accounting ApproachTo obtain cost annuity, it is necessary to transform thelong-run incremental costs incurred, including thoseresulting from investments, into an annuity. Anaccounting approach would have led to consideringyearly depreciation installments based on accountinglifetime of respective equipment, calculated accordingto fiscal criteria (linear, tapering, or accelerated depre-ciation).The cost annuity is determined here using aneconomic approach. Economic-oriented costs areassumed to be the effective tool for regulating industryentry and investments. It is assumed that entry orinvestment decisions are made on the grounds of prof-its or reasonable return on investment expected fromthe activity. In the proposed cost model, investmentcosts are converted into annual economic costs, asdescribed in appendix 1.

To proceed further with the discussion, we nowneed to introduce the cost of capital concept. Asdescribed in appendix 2, the cost of capital representsthe cost incurred by sponsors or promoters in mobiliz-ing financial resources.The cost of capital concept fac-tors in technology progress, economic, andcountry-specific risks.

A Bottom-Up ApproachTwo main alternative approaches can be used to esti-mate the long-run average incremental costs: the top-down model and the bottom-up model. These twoapproaches can be summarized as follows:• Top-down cost models rely on the accounting data

and allocate costs to different services on the basisof the causality correlation between the costs and

services. In some cases, top-down models are alsoimplemented with current costs.

• Bottom-up cost models imply the development ofengineering and economic models in order to cal-culate the costs of network elements used to pro-vide particular services, assuming an efficientoperator that uses best available technologies.

In principle, both methods should lead to the sameresult. In fact, this can only happen if the same assump-tions are made for operation efficiency and depreciation.The proposed cost model belongs to the second category.

The Working UnitsThe model has to determine unit interconnectioncosts. It is, therefore, necessary to select a metric sys-tem.Traditionally, the traffic flow through the core net-work is measured in minutes.The cumulated durationof calls, measured in minutes, is considered to be theimportant parameter in determining costs. However,some interconnection prices and some cost modelsalso take into account the number of calls.

In practice, network elements also handle the fol-lowing noninvoiced traffic: (a) the call set-up time, (b)the closing time, and (c) the time spent in handlingunsuccessful calls.The sizing of some network elementstherefore depends on the number of calls conveyed.This is particularly relevant for switching elements thathave to handle call attempts.1 The number of callsessentially continues to determine the cost incurred bythe temporal occupation of noninvoiced network ele-ments (waiting time and unanswered calls).The cost perminute is, by convention, the representative unit cost.

The Model Structure

The logical structure of the model is relatively simple:• The model begins by proposing a nomenclature of

network elements (nodes and links).• Each service uses these elements in different pro-

portions. On the basis of routing factors, the modelcalculates the total load supported in traffic minutesfor each element.

• The model sizes network elements specifically forthe transmission system, within the confines of thetopology the user chooses.


30

• The model then aggregates network elements costs.• The model finally calculates the interconnection

costs, depending on what network elements areused to supply the service.The network is made of network elements. Any

communication or any interconnection service (whichis a special form of communication) uses, on average, xtimes each of these elements (with x varying from 0 toa few units at a maximum). The x factors are calledrouting factors. This allows for the calculation of thetraffic flow per network element.The model calculatesthe investment cost incurred to satisfy the trafficdemand and derives respective operating costs.

The operating costs consist of two terms:• A cost subcomponent, proportional to investment

costs and reflecting the cost of maintenance anddirect operation of network element (spare parts,equipment section of preventive and correctivemaintenance, energy consumed).

• A cost subcomponent for the staff allocated tooperations.For small and spread-out networks, the staff cost

can hardly be evaluated as a percentage of investmentcosts. For the sake of accuracy, the user is requested tostate a reasonable number of staff required to run thenetwork.The model then uses unit labor cost input tocalculate the total of operating costs and divide itamong the number of network elements.

Obviously, investment costs are subject to detailedcalculations. Based on an estimate of the volume oftraffic to be handled by each network element, and theunit cost for each element, the model determines thetotal investment cost. To achieve this, the amount oftraffic assigned to each network element at peak time isused for its sizing. Using the unit investment cost pernetwork element, the model calculates the total invest-ment cost per element. The operating costs are allo-cated to the investment unit costs to obtain a globalcost per element. A unit cost per minute handled isthen obtained by dividing the global total by the total

volume of traffic handled.These costs per minute arethen added up to reflect the network elementsinvolved in handling a specific interconnection servicerequest.They are then adjusted by the hourly gradientapplied to retail services to obtain the final intercon-nection rate.

To sum up, the model reconstructs the costs of thenetwork at two levels:• At the level of nodes (switching elements), invest-

ment costs are structured in fixed and variable com-ponents according to BHE (business hour Erlangstransformed into 2 Mbps) and to the number ofsubscribers, respectively.

• At the level of links, calculations are done at twosublevels:– A transmission level at which electronic trans-

mission equipment (mainly on SDH rings) isdesigned and sized;

– An infrastructure level at which the link substra-tum is designed. Generally, the infrastructurescomprise three types of technologies: trenches,microwaves, and satellite links.

The trenches are broken down by geo-type (urban,suburban, and rural) corresponding to different bury-ing techniques (wrapped trenches, ducted, buried).Themicrowaves are characterized by the nature of theirmasts (light, medium, or heavy).

The links are sized by the traffic at peak timesexpressed in capacity (Mbps) and are sized to handlethe switched telephone network traffic, as well as leasedlines bandwidth.Assumptions regarding the sharing ofcertain elements of basic infrastructure with other net-works (for example, access networks) subsequentlyallow for the sharing of costs,which are not fully borneby the core network.

To sum up, the model computation modules are pro-vided in the following 21 sheets structured as follows:• A menu sheet for the user navigation.• Twelve sheets forming the core piece of the model

described afterwards.• Four sheets specific to mobile networks (whose

results appear on the general results sheet).• One sheet on the sensitivity of the model to some

parameters.• Three specific management sheets (two sheets for

the publication of fixed and mobile reports and a

Modeling Principles

31

Investment Operatingcosts costs

Attributable costsCommon costs

sheet to manage the two languages and the defaultvalues).The 12 sheets in the general model break down as

follows:• Four sheets to gather the assumptions.• One sheet to calculate the traffic and sizing of the

network elements.• Two other sheets to size the transmission and infra-

structures.• Three sheets to calculate costs (switching, transmis-

sion, and infrastructures).• One sheet to recapitulate the total costs (including

the common costs) and to calculate the unit costsper minute and element.

• One sheet to present results.The core of the model is synthesized in the follow-

ing diagram:

Applications of the Cost Model

In practice, the cost modeling exercise enables regula-tors and operators to improve their knowledge of theindustry cost structure and its efficiency frontier. Ofcourse, the challenge is managing the informationasymmetry that characterizes the regulator and theregulated firm relationships. The following are thethree main advantages that cost models can provide:• Providing a catalytic effect on cost information col-

lection and retrieval by both the regulator and theregulated firm. To run the model, the regulatordraws up a comprehensive list of required informa-

tion from the operators and sets up procedures forthe ratification of the information supplied (forexample, by means of accounting audits, establish-ing ABC, reviewing investment invoices).

• Focusing the benchmarking on a limited amount ofsensitive information used as inputs for the decisionprocess. As a result, the regulator shifts away fromstandard benchmarking of tariffs observed betweencountries to benchmarking key input or parame-ters.

• Providing the opportunity to the regulator toimprove on its information and knowledge aboutnetwork architecture and topology. With a betterunderstanding of the network structure, the regula-tor can refine procedures implemented to collectinformation from regulated operators. In fact, datacollection must be executed according to formaland recurrent procedures.In conclusion, it is important that regulators ensure

the confidentiality of the information collected fromoperators. Despite most national regulations that giveregulators inquiry powers, additional attention isneeded to ensure and preserve the confidentiality ofthe collected information. In general, newly establishedregulators do not yet have sound and effective safe-guard measures ensuring the confidentiality of infor-mation collected from operators. Unless such measuresare implemented, it is likely that regulated operatorswill remain reluctant to fully disclose their strategicinformation.

Like other costing methodologies, LRIC modelsalso have consequential limitations, which are relatedto the wide scope of assumptions that foster theirdevelopment. A system of check and control handledin collaboration with the industry can limit the arbi-trariness of the most sensitive assumptions.The follow-ing are likely to be among the most sensitive modelparameters that should be reviewed:• The selected network topology (number and loca-

tion of nodes).• The percentage of traffic at peak hour and other

demand data critical to sizing the network ele-ments.

• The traffic growth as predicted by the industry.• An estimate of the staff number that could be con-

sidered “efficient.”


32


DemandDemand


Traf c



Unit CostsUCosts

Totals Results

Hypotheses

Traf c

Sizing

Cost

Result

Tech FactRout

Capa Eltr

Costs Sw

tot results


Capa Intra

Capa

• Major cost drivers specific to the country and howthey impact the investment costs in comparisonwith international best practice.

• The proportion of joint and common costs.• The problematic decision to factor in TDMA con-

centrator systems in the increment leading to thecalculation of interconnection cost.A contradictory discussion on the model assump-

tions would most likely improve on the accuracy of therange of cost estimates provided. This accuracy willimprove as procedures prevailing to data collection alsoimprove. It will also improve the soundness of theargument against the theoretical criticism of LRICmodels.

Modeling Principles

33

Note

1. As a matter of fact, for a long time thenumber of call attempts during peak time(business hour call attempts [BHCA]) wasused to specify the overall processing capac-ity of switching exchanges. In view of theenhanced processing power of availableprocessors and the resulting decreasing costof memory, the number of calls processed atbusy hours is no longer a relevant criterionto effectively differentiate switching systems.Vendors now agree to consider the total du-ration of calls (measured in Erlangs) and thenumber of subscribers accessing the node asthe most important engineering parametersfor switching exchange design.

34

The cost model includes a friendly interface. It wasdeveloped as a Microsoft 2000 Excel file that containsseveral folders. The model depends on several macrocommands written in Visual Basic. In other words, theprogram cost model is less than 1 mega-octet in size;hence, it is easily portable.

After launching the program, the user can choosewhether or not to activate the macros links.Activatingthe macros is essential for the Menu sheet functions.

After the macros are activated, the user is asked toselect the language.There are two options: French orEnglish.The model is then presented with captions inthe selected language.

Once the language is selected, theCommand/Menu sheet is shown:

This sheet includes three sets of commands:• On the left is a column allowing access to the

Assumptions menu.• In the center is a column providing access to the

Options menu.• On the right is a column allowing the user to dis-

play or print out results and test their sensitivity.IMPORTANT: Save your working file in a

different name.

5 User Guide

User Guide

35

Filling in the Assumptions

The assumptions are entered on five sheets, which areaccessible via the Menu buttons in the left column ofthe Menu sheet:• Demand assumptions.• Network assumptions.• Routing factors.• Cost elements assumptions.• Specific assumptions for the mobile network.

The color code used to differentiate cells is as fol-lows:• Blue cells: exogenous assumptions to be filled in by

the user. These assumptions characterize the net-work and are essential for the modeling.

• Light green cells: default information that can bemodified.These assumptions can be shared by dif-ferent networks.

• Light blue cells: default values that can be modified.These default values are computed by the user andare, generally, presented in a table on the right.However, to run the model with provided defaultvalues, the user should avoid modifying the filledmatrix that contains default routing factors.

• Dark green cells: result cells, which under no cir-cumstances should be changed.

It is not advisable to attempt modification of themodel program or Visual Basic macros, unless one has agood knowledge of computer programming. Theentire range of sheets and all the formulas are accessi-ble.The user can follow a step-by-step overview of themodel’s computation process. However, modification1

outside the sheets that are accessible via the Menusheet (and the dark green boxes on these sheets) mayalter the model’s performance.

Demand AssumptionsThe Demand sheet is accessed via the Demandassumptions button on the menu (see below).

Traffic DataThe incumbent’s traffic has to be broken down into 11traffic categories.The total traffic in minutes has to beplaced in the first column. In the second column, theaverage call duration in minutes (with at least 1 or 2decimals) is provided by type of calls.The total numberof calls is then calculated in the third column. Finally,the predicted growth rate applied to the traffic volume(in minutes) that is used to measure the network isentered in the fourth column.

The traffic growth rate should remain moderate, toavoid major oversizing of the network.The cost calcu-

Margin for growth (%) Existing traffic over Number Average length Number of Numberincumbent’s network of minutes of calls successful calls of minutes

Local telephony 0Internet calls 0Long distance 0International* incoming 0* outgoing 0Calls to mobiles* domestic 0Calls from mobiles* domestic 0* international 0Interconnection* local 0* long distance 0Others (Kiosk. switched X25 etc.) 0Total 0 0

lation is done on a yearly basis. As a result, the modelconsiders the predicted increment of the traffic for theyear under review.

The following input is needed for the calculation:

First of all, the average answering time for success-ful and unsuccessful calls is taken by default at 15 and30 seconds.The percentage of successful calls is taken at75 percent.These values can be modified.

The percentage of the total traffic at the peak hour(business hour) is used to size the network’s capacity.This value is estimated by default at 0.04 percent (= 0.00040), but can be modified. It is considered thatthere are 335 peak days in a year (11/12 of the year),with on average 7.5 peak hours per day.This gives in all2,513 peak hours per year.The total peak hours onlyrepresents 1/2513 of the total annual traffic, hence 0.04percent. The user should note that this variable isextremely sensitive and exerts a direct impact on thesizing of the network.

The “gradient,” which is related to retail servicespricing schemes, should be calculated.The model pro-poses five pricing levels: a peak rate, a reduced rate foroff-peak hours, a reduced rate for weekend rate, andtwo other reduced rates to fit with more refined pric-ing schemes, if needed.

In the first column, the breakdown of the total traf-fic is entered as a percentage for each pricing level dis-cussed above. In the second column, the level for eachpricing level is entered in relative terms, with the peakhour rate being the reference. For example, if the peakhour rate is 60 cents and the reduced rate is 40 cents,we will enter 40/60 = 0.667 in the second box of thesecond column of the table.

The gradient is calculated in column 3 and will beused to calculate interconnection rates in these differ-ent pricing ranges.

SubscribersData on subscribers and their connection mode arerequested in the second part of this sheet. First, the net-work topology is filled as follows:

The first line of this table contains the total numberof nodes in each category.We will count:• RCUs: the total number of remote concentrator

units included in the network.• LSs: the total number of local exchanges equipped

with their own processing and command units. LSsmay also perform domestic or international transitfunctions.

• TSWs: transit exchanges, dedicated exclusively tohandling domestic transit.

• ISs: international exchanges, dedicated to interna-tional transit.

• CSs: central stations for TDMA systems, used toextend the network coverage to rural communities.2

• TSs: terminal stations in TDMA systems, used toconnect about 40 subscribers.


36

Traffic statistic Successful calls Unsuccessful calls

Average call set-up time inseconds (time to answer) 15 30Percentage of successful calls 75%Traffic in busiest hour of the year (as a percentageof the total) 0.00040

Retail rate Ratesgradient (ratio) Traffic (%) 100 = peak Gradient

Peak 100Off-peakWeekendOther 1Other 2Average 0

Node information RCU LS TSW IS CS TS Total

Number of nodes(total) 0Nodeslinked by satellite 0Other nodes not linkedthrough a SDHring 0Nodes on SDH rings 0 0 0 0 0

The second line of the table deducts the nodesaccounted for in the first line, and which are connectedthrough the domestic satellite system. In general, onlyRCUs or LSs are connected by satellite. The othernodes of the network are assumed to be connected toeach other through SDH rings. However, these ringsare not always rolled out for several reasons.The thirdline of the table deducts the nodes already accountedfor, and which are not connected by SDH rings.Thisrepresents an arbitrage in terms of efficiency that regu-lators should assess and decide upon.These nodes willbe connected by microwave, at a higher cost than thatof equivalent SDH rings.The last line then calculatesthe number of nodes interconnected by SDH rings.

The following table shows data relating to the sub-scribers’ connection:

The first two lines should provide the installedcapacity of the RCUs and LSs.This total appears onthe following line, beside which the total capacity ofthe TDMA systems is indicated. The two followinglines enable the user to enter the proportion of sub-scribers’ lines connected to RCUs and LSs. It alsoallows for access to the proportion connected to LSswith transit features. Doing that improves the accuracyof the routing factor estimates.

The table also includes information on the stockof subscribers (connected capacity), and differenti-ates those connected to RCUs from those connectedto LSs:• The first column contains the total number of net-

work subscribers.• The second column contains those connected to

nodes linked by satellite.• The third column contains those connected to

nodes that are neither linked by satellite or SDHrings.

• The fourth column contains the total number ofsubscribers connected by TDMA systems.

First of all, the user enters the number of 64 kbpsequivalent analogue or digital leased lines.The user alsoprovides the predicted growth rate to be taken intoaccount for network sizing. It should be rememberedthat all the links using the core network must be takeninto account:• Telegraph and telex lines.• Leased lines provided to private or public users

(banks, transporters, civil service departments,police, army).

• Leased lines provided to third-party operators(mobile networks).

• Leased lines used by the operator to set up datatransmission networks (X25, Frame Relay, IP net-works).

• Leased lines eventually provided to carry TV andradio programs.

User Guide

37

Form of subscriber connection

Percentage (%) o.w.of installed Installed TDMAlines: capacity systems

– Remote Concen- trator Units andrural switches

– Local switchesTotal 0o.w. % on local switch with transitcapabilitieso.w. % on local switch without transitcapabilities

Percentage of o.w. o.w. non- o.w.subscriber lines Used linked by linked by TDMAconnected to: capacity satellite SDH systems– Remote Concen-

trator Units andrural switches

– Local switchesTotal 0 0 0

Leased lines Number Margin for growth (%)

Analogue leased lines (64 kbits/s equivalents)Digital leased lines(64 kbits/s equivalents)

Number of local tariff zonesNumber of LS with transit capabilitiesNumber of TS interconnected to mobile networksRatio od traffic from TDMA subscriber to nonTDMA subscriberShare of direct long distance calls between LS% of direct routes betweenLS with transit capabilities

This is followed by additional information on refin-ing the default calculation of the routing factors:• Number of local pricing zones.• Number of LSs with a transit function.• Number of exchanges (LS-TS) where there is an

interconnection with mobile networks.• Ratio of the average traffic for a TDMA subscriber

relative to a non-TDMA subscriber (generallylower than 1).

• Proportion of the long distance calls between LSsthrough direct routes.

• Proportion of direct routes between LSs with atransit function.

Main Technical AssumptionsThis second (“Tech”) sheet containing complementaryassumptions concerning the network is accessed viathe Network assumptions button on the menu. Thissheet enables the user to enter important assumptionson the network structure.

Main Assumptions on Switching ExchangesWe start with a small number of assumptions wherebydefault values are proposed:

The first assumption relates to the percentage oftransit switches co-located with LSs so that the hostbuildings are not counted twice. If the network doesnot include transit switches, the assumption is not used.

The second assumption refers to the maximumnumber of nodes connected by an SDH ring. Thedefault value is 16 nodes.

Main Assumptions for Transmission LinksIn transmission, a link is the media over which infor-mation is conveyed between nodes.The information isconveyed through different routes that may exist overtransmission links connecting the nodes.The numberof routes that can be created between nodes onlydepends on the network security constraints.

The first step is to determine the number of 2Mbps channels that are needed to carry the traffic flow.Consequently, the user is asked to provide informationcharacterizing the transmission routes.

• In the first line, the load of the transmission ele-ments is provided as a proportion of the total capac-ity available.

• In the second line, information on traffic expressedin Erlangs is provided.The information provided above is used to com-

pute the size of the transmission network and thenumber of ports for each switching exchange.

The average length of transmission routes across allthe geo-types is then filled.The average route length isthe means of length routes between two nodes.

The following table portrays the transmission net-work structure for SDH systems. On the right, band-width capacity used for each link has to be provided.


38

RCU- LS-LS-LS TS to IS CS-TS

Utilization level of transmission elements (%) 60% 70% 80% 50%Erlangs per circuit 0.5 0.6 0.7 0.6Average length of transmission routes across all geo-types in meters

Distribution of traffic links

Mix of STM systems in a fully SDHtransmissionnetwork

RCU-LS LS-LS-TS to ISSTM 1STM 4STM 16STM 64Total = 100% 0% 0% 0%Note:These may be re-allocated to meet required capacity.

Co-locationof switches

Percentage of tandem switches co-located withlocal switches 50%Maximum number of nodes on a ring

Maximum number of nodes on a ring 16Utilization level of switching nodes (%) RCU LS TS IS

Used capacity (Erlangs) 95% 95% 95% 80%

User Guide

39

Routes are distributed according to their band-width needs. That makes it possible to size the elec-tronic transmission elements.These indications can beadjusted to the maximum number of authorized nodesper route. These calculations are carried out in theCapa ElTr sheet.

The following are additional default values used inthe computation of the network size:

The first indicator shows the average distancebetween installed microwave towers (in meters). Thedefault value is equal to 40 km.This length depends onthe relief, stretches of water, and so forth.

The second indicator shows the average distancebetween regenerators on SDH rings.The default num-ber is equal to 64 km.The load factor of leased lines istaken at 100 percent.

The diversity indicators make it possible to takeinto account the specificities of the real topology andtheir implications, with respect to an optimized topol-ogy retained by the model.

Main Assumptions on InfrastructuresThis part of the Tech sheet captures all the assumptionsneeded to determine the size of the infrastructures.

The first table captures the total length of infra-structures per type of technology:

The model distinguishes the total length (inmeters) by type of:• Trenches.• Microwave networks closing SDH rings.• Microwave networks supporting non-SDH links.

The cost distribution between LS-LS-TS and TS-TS is provided:

Trench specificities are provided: (a) wrappedtrenches in urban environment, (b) ducts in suburbanenvironment, (c) buried trenches in rural environment.The user is asked to fill in the distribution that is relevant:

Box 5.1 Erlang

An Erlang is a telecommunications traffic measurement metric.For example, if a group of users makes 30 calls in 1 hour and eachcall has an average duration of 5 minutes, the total traffic flow isexpressed as follows:

Minutes of traffic = number of calls x duration = 30 x 5 = 150 Traffic in Erlang = 150 / 60 = 2.5

Agner Krarup Erlang was born in Linborg, Denmark, in 1878. Hewas a pioneer of telephone traffic analysis. He proposed a formulafor calculating the proportion of callers served by an exchangewho are compelled to wait for their turn. In 1909, he published hisfirst result, titled The Theory of Probabilities and Telephone Conversa-tions. His work earned him worldwide recognition, as well as thatof the British General Post Office,which endorsed his formula.Heworked for 20 years at the Copenhagen Telephone Company untilhis death in 1929. During the 1940s, the Erlang became theaccepted unit for measuring telephone traffic.

Average length between microwave towers (meters) 40,000Distance between regenerators (in meters) 64,000Utilization level of leased lines 100%Diversity for STM multiplex and LTES(percentage) 15%

Other equipment

Transmission cross connects per IS 2Diversity for regenerators 2Number of repeater stations of TDMA systemsUtilization level of satellite transponders (%)

RCU- LS-LS-Length (meters) LS TS to IS Total

Total length of trench by transmission link(meters) 0Total length of microwave by transmissionlink (SDH closure) 0Total length of microwave by transmissionlink (non-SDH) 0Total (meters) 0 0 0

LS-LS TS-TS

Breakdown of transmission and infrastructure costs between single and double transit 75% 25%


40

Then, some complementary indicators related tothe cable network are required:

• The percentage of trenches (wrapped/ducts)equally shared with other networks, including theaccess network, the cable network, or a privatetransmission network. In general, fully buried cablesare dedicated to a single type of network.

• The percentage of trenches not equally shared withother networks. In general, in such circumstances,the core network only occupies about 25 percent ofthe trench’s capacity.This is given as a default value.

• The number of cables per duct (generally one).This is followed by assumptions concerning

microwaves.

How many SDH rings are closed by microwaves?Microwave links considered for closure operate at 155

Mbit/s. For non-SDH routes, the user is asked to fillin information on the number of routes per link typeequipped with microwave, and the distribution over34, 8, and 2 Mbit/s microwave.

Cost elements for microwave sites are discussednext:

• How many routes share the same microwave sites?What is the distribution between the core andaccess networks, or the TDMA network?

• How many links share the same antennae masts?– Light mast (lower than 40 meters)– Medium mast (between 40 and 60 meters)– Heavy mast (higher than 60 meters)Other costs shared between the core and the access

networks are discussed:

The following costs are attributed to the core net-work by default:• 50 percent of the costs for the RCU and LS site.• 70 percent of the costs of the site for terminal sta-

tions in TDMA systems.

Proportion of total length of trench in each geo-type (%): RCU-LS LS-LS-TS to IS

– urban (duct) – suburban – rural (buried)

0% 0% 0%

Trench sharing

Percentage of trenches that are shared by geo-type:

- urban - suburban - rural 0%

Percentage of shared trench attributable to conveyance 25% Cables

Cables per duct 1

Microwave

RCU- LS-LS-Closure of SDH rings LS TSW to IS

Number of routes

RCU- LS-LS-Non-SDH routes LS TSW to IS

Number of routesBreakdown per capacity34 Mbps8 Mbps2 Mbps

0% 0% 0%

% of shared masts ondifferent routesBackbone networkTDMA systemsPercentage of microwaveroutes served by RCU- LS-LS-different masts: LS TS to IS CS-TS

– Light– Medium– Heavy

Total 0% 0% 0% 0%

Site costs Access Core Other

Percentage of RCU sites attributed to service 50% 50% 0%Percentage of LS sitesattributed to service 50% 50% 0%Percentage of TDMA TS sitesattributed to service 30% 70% 0%

Percentage of site costsallocated to transmission (asopposed to switching) 25%Percentage of TDMA costsallocated to transmission (asopposed to switching) 15%

User Guide

41

The following are attributed by default to theswitching node:• 25 percent of site costs (and therefore 75 percent to

transmission).• 15 percent of TDMA system costs (and therefore 85

percent to transmission).

Main Assumptions on OperationFinally, this Tech sheet provides the number and thedistribution of staff between the core network, access,and other activities.

The user is asked to provide an estimate of thenumber of staff considered optimal by an efficientoperator.The total obtained is then distributed over theswitching, transmission, and infrastructure activities:

Main Assumptions on Routing FactorsThe RoutFact sheet is accessed by the Routing factorsbutton on the menu. This sheet includes 12 routingfactor matrixes.

A table showing the default routing factors calcu-lated on the basis of the information supplied in thetwo previous sheets is placed on the right. A similarblank table is presented on the left, allowing the user toenter his/her own computed routing factors if thedefault values calculated are unsuitable.

The first three tables are the routing factors for thefixed-line network, while the latter three are for themobile network. In each case, the first two tables indi-cate how many times on average each type of call useseach network element.The first information is for the“nodes” elements and the second is for the “links” ele-ments.The latter indicates how many times each typeof interconnection service uses these same elements(nodes and links) on average.

IMPORTANT: As soon as a blank table isfilled in by the user (even a single box), it will betaken into account by the model.Any table thatsubstitutes for the default table must thereforebe filled in completely. Some boxes can be leftempty, but all of the boxes that do not have anull value must be filled in.The user can specifya table without having to specify them all.

The first table relates to the nodes:

Thus, the first line should indicate how manyRCUs, LSs, and TSs are crossed on average by a localtelephone call and, subsequently, for all the types ofcalls according to their characteristics.

The second table requires the same information forthe different types of links, while adding leased lines:

Personnel (core network) Number

SwitchingTransmissionInfrastructureBreakdown RCU LS TS IS CS TS

Switching 25% 25% 0% 5% 15% 30%RCU- LS-LS- to

LS TS IS CS-TS

Transmission 15% 50% 5% 30%Infrastructure 15% 50% 5% 30%

Routing factors RCU LS TS IS CS TS

Local telephoneInternet callsLong distanceInternational

– incoming– outgoing

Calls to mobiles– domestic

Calls from mobiles– domestic– international

Interconnection– local– long distance

Others (Kiosk,switches X25 etc.)

RCU- LS-LS-Routing factors LS TS to IS CS-TS

LocalInternet callsLong distanceInternational

– incoming– outgoing

Calls to mobiles– domestic

Calls from mobiles– domestic– international

Interconnection– local– long distance

Others (payphone,ISDN, etc.)Leased lines

Finally, the third table gathers the same informationfor each interconnection service:

This third table covers both the nodes and links.

Main Cost Allocation AssumptionsThe Ucosts sheet is accessed via the Costs button onthe menu.

This sheet includes four major series of assump-tions:1. General assumptions on costs.2. Assumptions in regard to common costs, and apply-

ing to attributable costs calculated by the model.3. Assumptions in regard to the unit costs for each

network element for the fixed-line network.4. Complementary assumptions for the mobile net-

work.

General AssumptionsThese assumptions are important. First, they relate tothe currency in which the results will be presented andthe costs calculated. The name of this currency issought in the first line and then its exchange rate toeuros.

The model can be run with local currency or ineuros or U.S. dollars.An exchange rate is provided forthe conversion into local currency.The proposed costmodel uses the euro as the reference currency, but thisreference can be easily changed. Consequently, defaultunit costs are expressed in euros.

The model then requires three elements of supple-mentary information on the investment cost:

1. The first is the level of the customs and transit dutyapplied to imported telecommunications equip-ment. In general, customs duties can be very high,though, temporarily, exemptions are often grantedto operators as part of incentive schemes imple-mented in order to develop foreign direct invest-ment. In the proposed cost model, it is assumed thatall imported equipment is subject to custom duties.

2. The second is a market surcharge reflecting price dis-tortions related to the small size of Africaneconomies.This surcharge applies to imported equip-ment.Equipment prices in Africa are generally higherthan international prices. Because of the tiny size ofAfrican telecommunications markets, orders placedby operators are small in volume. Consequently,African telecommunications operators’ negotiationpower, vis-à-vis major vendors, is limited.

3. The third is related to transport and insurance coststhat are significantly higher for landlocked countries.These extra charges considerably increase the cost

of imported equipment. The model provides theopportunity to factor in this consideration whenreviewing the network’s costs. Salary and labor costs arein local currency.These costs are obtained by dividingthe staff expenditure incurred by the operator by thenumber of employees.


42

Inter- Local Single Double nationallevel transit transit Transit transit

SwitchingRCULSTSISCSTSTransmissionRCU-LSLS-LS-TSTS-TSto ISCS-TS

LocalModel currency

currency versus Euro Euro

NameExchangerate versus Euro 1,000Customs and transit dutiesMarketsurcharge TotalInsurance and freight 100%

Model currency Euro Local currency

Annual cost ofan employee

Options for costannuity: 1 1 annuity including price

trend 2 annuity without price

trend

Finally, the model requires that the user clearlyshow how cost annuities are computed. For furtherdetails on different methodologies, the reader isreferred to appendix 1. Two options are provided: (a)option 1 is to factor in the equipment price trend inthe computation; (b) option 2 does not integrate anyprice trend for the equipment. In general, the user willselect an option that takes into account the pricedevelopment of each network’s element.

The following are the key elements of the financialcalculations:1. Appendix 2 reviews the cost of capital calculation.

Information required to compute the cost of capi-tal is provided below.

2. The user introduces the margin of working capitalneeded to efficiently operate the network. This isprovided as a percentage of the total attributableand common over investment and operating costs.

The cost of capital resulting from the assumptionsdescribed above is then presented in a dark green cell.

In other words, the operator offering interconnec-tion services has two options to finance its investment:debt and equity.The gearing ratio captures the leverageexposure of the operator. Debt and equity are usuallypriced at different rates as their respective risks are dif-

ferent. Equity has a risk premium calculated on thebasis of the average return on investment observed inthe local stock market and a sector-specific factor (betacoefficient). Meanwhile, debt has a specific risk pre-mium factor.

Common Cost AssumptionsThe second part of the sheet deals with the commoncosts ratio.There is no strict standard in this area.Theapplicable common cost ratio does not necessarily haveto be the ratio derived from the operator’s books.Theratio should refer to those observed from efficientoperators. This is an area for which the regulatorshould conduct effective benchmarking.

The following are the default values provided bythe cost model:

Common investment costs include, for example,the vehicle fleet, other investment at the headquarters,and so forth. These costs are distributed among thefirm’s activities.

Unit Cost AssumptionsThe following unit cost assumptions are considered foreach network element:

User Guide

43

Level of gearing [D/(D+E)] 35%Risk free return 8%Average rate of return of the overall market 15%Risk premium on stock (beta coefficient) 0.80Risk premium on debt (spread) 2%Income tax rate 35%

Cost of capital (pre-tax) (%) 10.0%Working capital surcharge (%) 0.0%

Common costsattributed to in % RCU LS TS IS CS TS

Investments 5% 5% 5% 5% 5% 5%Operation 10% 10% 10% 10% 10% 10%

Common costs RCU- LS- TS- to attributed to in % LS LS-TS TS IS CS-TS

Investments 5% 5% 5% 5% 5%Operation 10% 10% 10% 10% 10%

Operating User Default costs as a

Equipment input value Installation Evolution Scrap value percentage of Price local Equipment costs Asset Price of as a % of equipment local currency Price (% of capita life trend Capital* equipment capital

Equipment X currency FAB Euros FAB costs) (years) (%) price capital cost cost (%)

Equipment XFixed cost of equipment — 147,000 10% 11 –8% 0.92 1% 3.0%Site cost of switch — 200,000 10% 38 –1% 0.99 11% 5.0%Cost per line — 10 10% 12 –8% 0.92 1% 2.0%Cost per trunk — 1,500 10% 12 –8% 0.92 1% 2.0%


44

1. 1st column (column C in the model): contains theresult of the unit cost calculation.The user shouldavoid changing this information.This value deriveseither from input placed in the third column(expressed in euros) or from the value entered bythe user in the second column.The following cal-culation is done :a. On the basis of the default value expressed in

euros (third column), the exchange rates and thethree rates stipulated for imported equipment(customs, market surcharge, and transport) areapplied;

b. On the basis of a value entered by the user (entryin column 2, column D, and takes precedenceover the default value) in local currency, and twoof the three rates stipulated in reference toimported equipment (customs and transport, butno market surcharge) are applied.

2. 2nd column (D): data entered by the user as a unitcost that is more suitable than the default cost. Datain national currency takes precedence over thedefault cost.

3. 3rd column (E): default value of the unit costexpressed in euros.

4. 4th column (F): equipment installation cost (as apercentage of the capital cost). This cost includespossible engineering costs (survey, planning,design), costs for monitoring and possibly inspect-ing the manufacturing process, the installation costsper se, the costs for system testing, and costs fortraining (on site or abroad).This column contains adefault value, which can be modified by the user.

5. 5th column (G): an economic lifetime and not anaccounting lifetime. This duration is implicitlyadjusted if cost annuity is done according to option1 (price trend).This lifetime is determined by theequipment’s operating longevity, availability of spareparts, and so forth.This column contains a defaultvalue that can be modified by the user.

6. 6th column (H): price development trend.This is thetrend for the long-term development of equipmentcost (+ or – x% per year). The column contains adefault value that can be modified by the user.

7. 7th column (I): contains the cost of capital and ofthe price development.This value is used to com-pute cost annuity when option 1 is selected.

8. 8th column (J): residual value of the equipment atthe end of its lifetime. The residual value isdeducted from the capital cost used in calculatingthe constant economic annuity. It is primarily sig-nificant for buildings and sites for which the landkeeps a high residual value.This column contains adefault value that can be modified by the user.

9. 9th column (K): operating cost incurred by theequipment as a percentage of the capital cost.Thiscost is the operation-maintenance cost incurreddirectly by the equipment, and does not includestaff or labor costs. It comprises spare parts, repaircosts, and expenditure for equipment consumables(energy).This column contains a default value thatcan be modified by the user.The number of unit cost items amounts to 68 and

is grouped below as follows:• Eighteen headings for switching.• Twenty-four headings for transmission (SDH elec-

tronics, microwaves, towers).• Nine headings for infrastructures (cables and ducts).• Four items for other costs (including one related to

mobile networks).• Thirteen cost headings specific to mobile networks.

Some of the headings deserve the following specificcomments:• Locally supplied equipment or local construction

(buildings, civil engineering) are not subject toimported expenses (custom duties and taxes).

• Cost heading related to Domsat repeaters; leasedlines for mobile networks belong to operating cost.

• For some investment, the model does not identifyspecific cost items.These investments are assessed asa percentage of investment cost implied by a basketof items. The model selects four baskets for thesecomplementary investments:– Switching investments: this item includes every-

thing needed for the operation of basic services,excepting value-added services, and does notshow price entered in the equipment. Thisapplies to: SS7 signaling, signal transfer points(generally included with the switches), synchro-nization, centralized network management sys-tems and, possibly, training equipment.

– Transmission investments: this includes every-thing related to centralized management, alarm

User Guide

45

checking, and testing and installation equipmentthat is not included in material supplies alreadyaccounted for.

– Infrastructure investments: testing, layout, instal-lation equipment.

– Mobile investments: items equivalent to themobile networks that are not included in theunit prices for equipment.

In general, the cost of network elements is dividedinto two categories of cost item:1. The fixed cost of equipment (processing unit of a

switch, racks, management bays). These costs varywith capacity (subscribers, BHE, number of fibers).

2. The cost of the site where the equipment isinstalled.The site costs deserve clarification. In general, the

site includes the following cost elements:• Land acquisition cost.• Land development and building costs (fencing,

access road).• Specific costs for fitting out buildings (technical

floors, air conditioning, protection against lightningand fire, security system).

• Energy cost.A telecommunications building often contains:

• Switching and transmission equipment.• Equipment for the core network or for the access

network, and even for other activities (for example,accommodation for public telephone boxes, a salesbranch).Consequently, site costs should be shared out

among the various activities. The model handles theallocation of these costs in the Tech sheet, which hasalready been reviewed. The cost for sites (like civilengineering work) is considered to be a locally pro-vided service. Hence, it is not subject to the three ratesspecified in the model (customs, market cost overrun,transport). The same applies to costs of masts andmicrowave sites.

Assumptions for the Mobile NetworkSeveral assumptions in the preceding sheets are used tocalculate interconnection costs for mobile networks.This, in particular, is the case with the Cost sheetreviewed earlier. Nonetheless, specific demand andtechnical elements are required to determine mobile

network costs.These are entered in the Mobile sheet,which the user can access via the Mobile assumptionsbutton on the Command (Menu) sheet.

The sheet is organized into three major sections:1. Demand assumptions.2. Subscriber and network assumptions.3. Operating cost assumptions.

Demand AssumptionsThe user can enter aggregate traffic elements carriedby the mobile network, similarly to what was done forthe fixed-line network:

Demand and network elements for mobile are, ofcourse, different:

Margin for growth (%)Existing Number Average Number Number traffic over of length of success- of mobile network minutes of calls ful calls minutes

OutgoingInternal 0to mobile network 0to fixed network 0Incomingfrom mobilenetwork 0from fixed network 0Other (CRM…)Other (CRM…) 0Total 0 0

MSC BSC BTS

% of internal calls using the same… 80% 60% 20%Successful Unsuccess-

Traffic Statistics calls ful callsAverage call set-up time in seconds (time to answer) 15 30Percentage of successful calls 77%Traffic in busiest hour of the year(as a percentage of the total) 0.0005Retail tariff gradient Traffic (%) Tariffs Gradient(ratio) 100 = peakPeak 100off-peakWeekendother 1 other 2Average 0


46

In addition, the model indicates the proportion ofinternal calls that terminate in the same mobile switch-ing center (MSC), base station controller (BSC), andbase terminal station (BTS) area.

Assumptions for the Mobile NetworkThis section contains information that specificallycharacterizes the mobile network and is required forcost determination:

1. Capacity installed on MSCs and the number ofsubscribers:

2. Equipment utilization levels:

3. Node information:

4. Data on BTSs:

The average number of calls transmitted by trans-mitter/receiver (TRX; this is the radio equipment thatmanages transmission at base station level) is betweenfour and eight.The number of TRXs per base stations,and the number of sectors by the BTSs and the num-ber of BTSs, are computed and entered by the user.These values are divided on the right by the number ofBTSs to ensure that the average values obtained areactually within the technical ranges accepted.

5. Data on the transmission network of the mobileoperator are entered when these links can be leasedfrom the incumbent fixed-line operators or fromany other licensed operator or are rolled out by themobile operator (owned microwave or fiber-opticlinks).

The user enters the type of link, the number ofroutes, length of routes, and distribution by capacity.

6. The breakdown of the number of masts is entered,thus:

Number

Installed capacity (MSC)Subscribers (post-paid)Subscribers (prepaid)

Node information MSC BSC BTS

Number of nodes (total) BSC co-located with MSC

BTS data Total per BTS

Number of communications per TRX 4 < x < 8

Total number of TRX (Full duplex channel) 6 < x < 16

Number of sectors (cells) 1 < x < 6

Number of sites < 1

Utilization level ofswitching nodes (%) MSC BSC-BTS

Used capacity Erlangs 95% 80%

MSC-MSC MSC-BSC BSC-BTS

Utilization level of transmission elements (%) 90% 70% 60%Erlangs per circuit 0.5 0.6 0.5

Transmission network

MSC- MSC- BSC-Microwave MSC BSC BTS Total

Number of routes 0Total length (m) 0

Break- Break-down down

Breakdown of total of totalper capacity routes length155 Mbps34 Mbps8 Mbps2 Mbps

Transmission MSC- MSC- BSC- MSC- MSC- BSC-electronics MSC BSC BTS MSC BSC BTS

155 Mbps 0 0 034 Mbps 0 0 08 Mbps 0 0 02 Mbps 0 0 0

0 0Total number of mastsPercentage of different masts:– on roof– Light– Medium– Heavy

The sizing of electronics equipment per link isdone automatically. If the breakdown appearing in cellsF76:H79 (model) does not match the actual situation,the user can enter the relevant data in cells C76:E79.The total number of physical masts should be enteredin cell C81.

The model then requests a breakdown of masts inthe following categories: (a) masts installed on roofs, (b)masts requiring light masts (<40 meters), (c) medium-sized masts (between 40 and 60 meters), and (d) heavymasts (>60 meters).

7. Leased lines used by the mobile network:

The model requires differentiating urban andnonurban leased lines. The user is requested to enterthe number of leased lines per type and per length.

Assumptions on Operating CostsThe Mobile sheet includes data needed to calculate thelabor costs and their breakdown:

The second part of the Cost sheet includes theassumptions required for common costs allocation.Data provided by default can be modified by the user.

The Results Sheets

After entering all the inputs and assumptions, the usercan display, print, or test the sensitivity of the model’sresults.These buttons activate four sheets:1. A Results sheet is displayed.2. The user can print the model’s results by selecting

the fixed-line or the mobile network simulation.Each report includes the results and assumptions.However, the intermediary computations are notprinted, though these can be displayed in differentspreadsheets.

3. The user can test the results’ sensitivity to predeter-mined parameter variations.

The ResultsThe simulation results for fixed-line and mobile oper-ator are presented in Print sheets. Each simulationincludes:• Costs per minute for each network element.• Interconnection services costs obtained from net-

work elements costs as follows: network elementscosts that are respectively multiplied by corre-sponding routing factors to derive interconnectionservice costs.For the fixed-line network, the cost model provides

the option to add rural TDMA systems costs to theinterconnection service costs.The model provides twointerconnection cost estimates. The first estimateincludes TDMA systems costs, while the second onedoes not. Depending on universal access policies, theregulator will decide whether to account for TDMAsystems costs.

User Guide

47

Leased lines MSC- MSC- BSC-(2 Mbps equiv.) MSC BSC BTS

Urban– Number

Nonurban– Number– Total length (km)

Personnel (network operation) Number

SwitchingTransmission

Breakdown MSC BSC BTSSwitching 25% 25% 50%

MSC-MSC MSC-BSC BSC-BTSTransmission 5% 15% 80%

Common costsattributed to in % MSC BSC BTS

Investments 5% 5% 5%Operation 10% 10% 10%

Common costsattributed to in % MSC-MSC MSC-BSC BSC-BTS

Investments 5% 5% 5%Operation 10% 10% 10%

Cost per Switching minute

RCULSTSISCSTSTransmissionRCU-LSLS-LS-TSTS-TSto ISCS-TS


48

The costs are expressed in the currency selected inthe unit cost sheet.They are converted into euros forcomparison purposes. These costs are average costs.Theyare then converted into costs according to time-of-dayslots with the gradients specified in the demandassumptions:

For the mobile networks, the results provide thecost per minute for network elements and the cost ofinterconnection services. A distinction is madebetween call termination and origination, wheneverjustified by the routing factors.

Local Single Double Transit 0 level transit transit Transit international

Interconnection rates (with TDMA systems)Interconnection rates (without TDMA systems)Africa “Best current practice” peak rates(Model Currency)Europe “Best current practice” peak rates (Model Currency)

Local Single Double Transit Euros cents level transit transit Transit international

Interconnection rates (with TDMA systems)Africa “Best current practice” peak ratesEurope “Best current practic” peak rates(Euros cents) 0.5 – 0.9 0.8 – 1.5 1.5 – 1.8 0.3 – 0.7

Time of day charging(Modelcurrency)

Transit With TDMA Local Single Double inter-systems level transit transit Transit national

PeakOff-peakWeekendOther 1Other 2

Switching Cost per minute

MSCBSC-BTSBTSTransmission

MSC-MSCMSC-BSCBSC-BTS

0 Originating Terminating

Interconnection ratesEuro cents

Interconnection rates

Time of day charging Originating Terminating(Model currency)PeakOff-peakWeekendOther 1Other 2

User Guide

49

Printing ReportsPrinting can be launched from the Menu sheet:• Results and assumptions for the fixed-line network.• Results and assumptions for the mobile network.

When launching the printer, the user is asked toprovide a name for the printed simulation.When pro-vided, the name is stated at the top of the printedreport for identification purposes.The statement is notcompulsory.

The document is printed using a print previewfunction that enables the user to change the format andadapt it to the available printer, if necessary. Printinggenerally requires eight pages.

The printer selection box is called up by clickingon the Print button on the upper bar:

The printer and printing parameters can then bespecified.

The same procedure applies to printing the mobilenetworks report.

Testing the Sensitivity of the ResultsOnce an initial determination of the interconnectioncosts is obtained, the sensitivity of the results can betested. The user can perform the sensitivity analysisthrough the Sensitivities button on the Menu sheet.

A snapshot of the sheet is provided below:

To test the model sensitivity, captions for six prede-termined parameters are provided.The user can enterinput as described below:• Traffic at peak hours (as a percentage of the total

traffic) (cell C31 of the Demand sheet).• Total length of the trenches (sum of cells C48 to

E48 of the Tech sheet).• Total staff (sum of cells C105 to C107 in the Tech

sheet).• Average annual cost of an employee (cell E16 of the

Ucosts sheet).• Market surcharge ratio (cell C12 of the Ucosts

sheet).• Level of gearing [D/(D + E)] (cell C21 of the

Ucosts sheet).The user can modify these variables, either by

entering a relative variation expressed in percentage(total length of trenches, staff, annual cost of anemployee) or by adding to or subtracting from theoriginal value, which is already a percentage (peakhour traffic, market surcharge, level of gearing).


50

The modification can be made directly by enteringa value in the cell assigned for new value; the user canincrease or decrease the variable value by utilizing thevertical “cursors.”

The simulation is run by pressing the Test buttonwhen the new values are entered.

The original values and the new sensitivity param-eter values appear in columns E and F of an Excelsheet.The new results can be compared with the oldones, as shown in the following table:

The user can iterate several sensitivity tests beforeconfirming and saving the resulting values.This is doneby pressing the Set sensitivities button, which opens adialogue box that requests the user to either validatethe modifications made or to cancel them.

If the user decides to keep original values, themodel’s assumptions are not modified. Conversely, ifthe user decides otherwise, the model’s assumptions aremodified accordingly.To be able to keep both originaland new values, the user must save respective files indifferent names.

A : Interconnection ratesB : Interconnection rates with TDMA cross-subsidization

Before changeSingle Double International

Local level transit transit Transit transitA Old valuesB

After changeSingle Double International

Local level transit transit Transit transitA New valuesB

ChangeSingle Double International

Local level transit transit Transit transitA DifferencesB

User Guide

51

Managing the Model• Users are advised to save completed simulations

with suitable names.The user can clear the model’sassumptions or restore the default values of themodel.

To run a simulation with the model, the user must:1. Enter the different assumptions or parameters

required. If needed, the user can also modify thedefault values provided.

2. After providing all the input, the user can displaythe results.

3. Test the sensitivity of these results to variations ofmain or key parameters.

4. Print out the reports, including the simulationresults, along with the underlying assumptions.

IMPORTANT: Whenever a logical problem(such as division by zero) arises, or there is anyinconsistency among assumptions, the modelwill not compute the interconnection costs. Forexample, sizing the capacity of SDH ringsassumes that all input provided by the user isconsistent with the traffic that will be conveyed.In this particular case, the user has to ratify theassumptions step by step. Doing this, he or shecan either modify the capacity of the SDH sys-tems (cells C26 to E31 of the Tech sheet), orreview the calculations done by the Capa EITrsheet.

Notes

1. This includes the formulas as well as thelayout (addition or suppression of lines orcolumns, and so forth).2. Thus, a TDMA system connected to anexchange that included a capacity of 512subscribers would count as two CSs.

52

6 Operations of the Cost Model

This section presents the proposed model in greaterdetail.

The model was developed using Excel 2000 andVisual Basic.All intermediary calculation sheets are inExcel 2000.

The model’s synopsis structure is shown below:

Assumption sheets are described in the user guide.This section further spells out how the intermediarycomputations are implemented by the model.

Basic Principles

The model builds on the breakdown of the core net-work, as discussed in previous chapters, into networkelements. Eleven network elements are identified: six

different types of nodes and five different types of links.These elements are extensively reviewed in chapter 2of this guidebook.

In terms of sizing, LS-LS and TS-TS links are bun-dled under the same LS-LS-TS caption to take intoaccount the fact that subscriber exchanges also offertransit features in Africa.As discussed in chapter 2, fewAfrican networks have dedicated transit exchanges.However, when calculating the interconnection costs,the model splits these costs into two categories: LS-LSlinks and TS-TS links.

To size the various elements, the model uses totaland the peak hour traffic information.This is done bythe Capa sheet.The network size is calculated for:• Transmission electronics equipment (Capa ElTr

sheet).• Infrastructures (Capa Infra sheet).

After the sizing is done, the model calculates thenetwork costs for each element:• The nodes comprise the subject of a cost break-

down on the Costs Sw sheet.• The link costs are broken down on two sheets:

– The transmission costs are calculated on theCosts Tr sheet.

– The infrastructure costs are calculated on theCosts Infra sheet.

Total traffic and costs (including the common costs)are consolidated before the model determines the unitcosts (Tot sheet). The Result sheet then presents unitinterconnection costs according to routing factors.

DemandDemand


Traf c



Unit CostsUCosts

Totals Results

Hypotheses

Traf c

Sizing

Cost

Result

Tech FactRout

Capa Eltr

Costs Sw

tot results


Capa Intra

Capa


Operations of the Cost Model

53

Logic of the Intermediary Spreadsheets

This section describes in detail the intermediary calcu-lations.

Capacities (Capa Sheet)This sheet calculates the total traffic and the peak hourtraffic using the information on volume of traffic inminutes, as entered by the user:a. The volume of traffic is adjusted to factor in non-

billed traffic, which is costly (time needed to setup the successful calls. Column C: number of calls* average time to answer successful calls in sec-onds /60).

b. The resulting value is adjusted with the time con-sumed to convey unsuccessful calls (column D: col-umn C + number of unsuccessful calls * averagetime to respond to unsuccessful calls/60).

c. This is then adjusted with the predicted trafficgrowth rate, as indicated in the Demand sheet.

d. As a result, the adjusted volume of traffic (invoicedand noninvoiced with a growth rate) is brokendown by network element (four nodes and threelinks, excluding the radio concentrator equippingbase stations of TDMA systems), based on routingfactors provided in the RoutFact sheet.

e. The model provides the volume of traffic in min-utes supported by each network element.

f. The total traffic obtained is converted into Erlangs(BHE) and used to size nodes and links.This infor-mation,depending on the charge factor of each ele-ment (percentage of occupation), is adjusted toensure traffic fluidity.

g. In the case of links, these BHEs are transformedinto Mbit/s.

h. The same calculation is applied to leased lines thenumber of which is converted into Mbit/s andadjusted with the predicted demand growth rate.

i. The core network’s capacity is adjusted accordingto resources required by the public switched teleph-ony network and leased lines traffic.

Transmission Capacities (Capa ElTr Sheet)This sheet breaks down the transmission equipment,which includes add-drop multiplexers (ADMs) andtermination equipment for SDH rings (STM 1, 4, 16,and 64).

a. The capacity of transmission elements is calculated.b. An adjustment factor is then applied to oversize the

capacity.The adjustment factor is taken equal to 1,by default. It can be adjusted in line 10, if excep-tional factors that apply to certain types of linkshave to be taken into account.

c. The model then calculates the proportion of thelink’s capacity not handled by SDH links from sub-scriber connections, which are not connected toSDH links. In the case of international transit, themodel assumes that links to the international transitexchange are point-to-point SDH. If this is not thecase, the user should, herein, indicate the propor-tion of these links that do not use SDH technology(cell E11).

d. The capacity (in Mbit/s) that has to be served inSDH, in Domsat, and in non-SDH terrestrial linesis then derived (lines 12 to 14).

e. The model then considers the number of nodes andthe positioning of the nodes with respect to links(lines 17 and 18), as well as the mix of SDH systemsfor the various categories of links (lines 27 to 30).Three types of links are served by SDH technolo-

gies: RCU-LS links, LS-LS-TS links, and links to theIS. For each, the model calculates the number of ringsneeded.The algorithm builds on a former one devel-oped by Europe Economics, and takes into account themaximum number of nodes per loop as entered in theTech sheet (cell C10).

The simulation carried out assumes that RCUs arelinked to LSs through SDH systems. LS-LS and LS-TSlinks are implemented through an upper-level SDHring.The ring’s capacity is sized according to the flowof traffic between its nodes. It is assumed that the min-imal number of nodes to justify a ring is three. Thetraffic capacity that can be handled per ring is deter-mined according to the following factors (STM 1, 4,16, and 64):• The number of rings constituting the core net-

work.• The average number of nodes per ring.• The number of physical routes per ring (number of

nodes + 1).• The number of ADMs and termination multiplex-

ers (MUXs) (2 * number of rings + total number ofnodes).

• The number of regenerators.


54

Links to ISs are assumed to be SDH point to pointand are, therefore, much easier to size. The followingbox summarizes the algorithm used by Europe Eco-nomics’ model to size transmission links.

Infrastructure Capacities (Capa Infra Sheet)This sheet sizes the infrastructures needed to roll outthe core network transmission links. In general, thereare two types of infrastructures:

• Trenches and ducts.• Radio links.

The trenches are, themselves, differentiated in threesubcategories, according to their location:• Urban areas: wrapped ducts.• Suburban areas: ducts.• Rural areas: buried cables.

In the proposed model, aerial cables are not usedfor conveying traffic in the core network.

Box 6.1 Sizing of the SDH Network (example of RCU-LS routes)

The transmission equipment required is evaluated by type of link.• Capacity required in Mbit/s = total capacity required in Mbit/s (adjusted) taken in the assumption section of the same sheet.• Distribution of the required capacity (Mbit/s).STM1 = required capacity * total STM1 system mix (the two data items are taken in this sheet) (A)STM4 = idem (B)STM16 = idem (C)STM64 = idem (D)• Number of nodes = number of RCUs taken in the same sheet, assumptions section.• Total number of nodes per STM1 capacity (STM1 share in the systems total by the number of nodes).

• Same formula for the STM4, STM16. and STM64 systems.• Same formula for the STM4, STM16. and STM64 systems.STM1:

Column 2: the preceding result * by the maximum capacity of a piece of SDH equipment (same sheet in assumptions).The same procedure is applied to STM4, STM16, and STM64 systems.• Number of extra nodes needed to have at least 3 nodes per ring: logical function.

capacityrequired in STM A

capacity required for all STM A B C Dnumber

of RCUs

1( )*

− − −( )∑

IF

number of nodes per capacity for STM1maximum number of nodes on a ring ("technical" sheet)

capacity required for STM1maximum capacity of a piece of SDH equipment

;

then take the first term;

else take the second.

⟩

IF

number of RCUs 3 * rings per capacity ; then marked " check" ;

else several cases are possible represented by logical functions:

IF

3 * (total number of STM1 rings total number of nodes(RCU)) 0;

then take this figure IF 3 * the same thing for STM4 ; IF 3 * the same thing for STM16 ;

IF 3 * the same thing for STM64 ; else 0

⟨

⟩+ ( ) + ( )

+ ( )

∑

–

(Continued on page 55.)


55

Box 6.1 Sizing of the SDH Network (example of RCU-LS routes) (continued)

• If we obtain “check” for the preceding calculation, i.e., if the number of RCUs is under 3 * the sum of the rings per capacity we thenrecalculate the total number of rings per STM1, STM4, STM16, and STM 64 capacity using an IF logical function.

• We recalculate the total number of nodes per capacity (STM1, STM4, STM16, and STM64) adjusted by 1 as a logical function again asa function of the calculation of the number of extra nodes needed to obtain at least 3 nodes per ring.

For example, for STM1 (the same formula is applied for the other capacities):

• Reallocation of nodes on rings by capacity.Column 1: for example for STM1 systems (the same formula is applied for other capacities):

Column 2:

• The total number of nodes by adjusted capacity 2:For example, for STM1 (the same formula is applied for other capacities):

• The average number of nodes per ring.For example, for STM1 (the same formula is applied for STM4, STM16, and STM64):

• Average number of physical routes per STM1 =

• Average number of physical routes per STM4 ring = same formula as STM1.• Average number of physical routes per STM16 ring = same formula as STM1.• Average number of physical routes per STM64 ring = same formula as STM1.

IF

C39(cell number) " check" ; then take a figure in the recalculated previous table i.e.

number of nodes * number of STM1 rings / total number of rings of all capacities;

else take a logical function IF

3 * total number of rings per capacity total number of nodes (RCU) ;

then take 3 * the total number of rings per capacity; else take the total number of nodes

− =

⟩

,

IFC39 " check" ; then take the number of nodes above;

else IF number of nodes 3 * number of rings 0; then take this number ; else take 0

=

( ) ⟩( )

–

IFC39 " check" ; then take 0;

else take (column1 STM1/ column 1) * the number of extra nodes needed to obtain 3 nodes per ring

=

∑

IF

C39 " check" ; then take the adjusted rounded number of nodes 1; else

IFthe adjusted number of nodes 1 total number of rings, take the rounded total adjusted number of nodes 1

else take the total adjusted number of nodes 1 preceding reallocated nodes column 2

==

;

–

SI

C39 " check" ; then IFtotal adjusted number of nodes by capacity 2 0 then 0 ;

else the total adjusted number of nodes by capacity 2

adjusted total number of rings

;

else IF total number of nodes by adjusted capacity 2 0; then 0; else adjusted total number of nodes by capacity 2

total number of rings

==

=

(Continued on page 56.)

IF(average number of nodes per ring = 0; then 0; else take the average number of nodes per ring +1)


56

TrenchesThe first table recapitulates the length of trenches bygeo-type (lines 7 to 10). Then the breakdown oftrenches is implemented as follows:1. Nonshared trenches (with other networks) are

identified and corresponding costs allocated to thecore network (lines 15 to 18).

2. Shared trenches are identified (lines 20 to 23).3. The proportion of shared trenches costs that are

attributable to the general network is then derived.4. The model adds directly attributable costs (non-

shared trenches) and indirectly attributable costs(proportion of shared trench costs) to obtain thetotal cost of trenches allocated to the core network.The length of cables is calculated in terms of the

number of cables declared per duct (1 by default, cellTech C70).The length of fibers required is calculated. Itis equal to the average length of the transmission routes

(all geo-types considered [Tech line 22]) multiplied bythe number of physical routes taken from the Capa ElTrsheet (route between two ADMs).The total obtained isthen multiplied by four, assuming that each route isserved by four fibers in general. By dividing the lengthof the fibers by the length of the cables, the averagenumber of fibers per cable is obtained. And, from thislatter information, the model determines the minimumfiber capacity of cable (6, 12, 24, 36, 48, and 96 fibers).

Microwave NetworkTwo cases are envisaged:• A microwave link is used to close an SDH ring.This

is usually justified when it is too costly to lay downa cable (stretches of water, mountains, crossing for-eign countries).

• The transmission network has not been upgradedto SDH technology.

Box 6.1 Sizing of the SDH Network (example of RCU-LS routes) (continued)

• Total number of physical routes:

• Number of termination systems = total number of physical routes (above) *2.• Number of multiplexers (Gateway MUX and ADM) by capacity:For example, for STM1:

• Number of regenerators:

The rounded function up (z,0) rounds the result of the ratio z upwards to the next integer.

IFC check39 ==

( )

" " ; then sommeprod (average number of physical routes * total number of rings by adjusted capacity for b98

"check"; else sommeprod (average number of routes * total number of rings by capacity)

The "sommeprod" function gives the sum of products of corresponding elements for several matrices :

sommeprod matrix1;matrix2;matrix3;...

IF C check39 = =( )+

" " ; then 2 * the total number of rings adjusted to take account of B98 "check"; else 2 * the total number of rings

the number of nodes by adjusted capacity 2

roundupRCU - LS average distance (" technical" sheet)

average distance between regenerators ("technical" sheet)diversity of regenerators (" technical" sheet)

* total number of physical routes

−, *0 1

The closure of SDH rings is done with 155 Mbit/ssystems. The non-SDH links are only recommendedfor narrow band systems (bandwidth inferior to 155Mbit/s).The non-SDH links are therefore dedicated toconnect areas or localities for which SDH systems arenot viable.Three capacity levels are retained: 34, 8, and2 Mbit/s.

For each type of link and technology, the modelcalculates:a. The number of radio systems from the number of

routes.b. The length of the networks and the average dis-

tance of leaps.c. The number of pylons and the breakdown of these

pylons by size (light, medium, and heavy).d. The necessary antenna equipment.

CostsThe costs are calculated on three distinct sheets: onefor the switching costs (Costs Sw), one for the trans-mission costs (Costs Tr), and one for the infrastructurecosts (Costs Infra).

At the head of each sheet, the model recapitulatesthe size and number of equipment planned, and infor-mation needed to allocate cost of shared equipment.The investment costs per equipment type are calcu-lated as follows:

Investment = volume required * unit cost *(1 + installation cost (in percent)) *(1 – residual value/(1 + life span) ^ (capital cost))

An investment annuity is calculated as follows (discount rate with or without incidence of theequipment price trend):

Annual repayment = investment / phi (life span,discount rate) where phi is the annuity function shown in appendix 1.

The final step is to calculate operating costs attrib-utable to each network element.

These sheets are organized as follows:

Switching Costs (Costs Sw Sheet)This sheet calculates the costs for the six types of nodeas shown below. Five cost headings are retained. Notethat these headings are not relevant for all the nodes.

• There are no subscriber line costs for transitswitches (TS and IS).

• There are no site costs for central stations of TDMAsystems (co-localized with the switches or the RCUs).

• The cost of subscriber cards is included in cost perline.

• There is no cost for trunk in radio concentratorequipment.

• The number of equipment is derived from theDemand and Capa sheets (for two Mbit/s ports).

Transmission Costs (Costs Tr Sheet)The transmission cost calculation is made as follows:


57

RCU LS TS ISI CS TS

Fixed costs of the switchSite cost ofthe switchCost per lineCost per trunk Other relatedcosts

RCU- LS- toLS LS TS-TS ITS CS-TS

STM 1STM 4STM 16STM 64RegeneratorsDigital cross connectsLine terminationsystems STM 1Line termination systems STM 4Line terminationsystems STM 16Line terminationsystems STM 64TDM relay stationsDomsat centralstationDomsat localstationSwitching sitesOther related costs

The cost headings, mentioned above, do not applyto all transmission links. For example, in the Domsatnetwork, the central station is accounted for at thetransit level, while local station costs are accounted forat the level of local exchanges. For operating costs, therepeaters and staff costs are added.

Infrastructure Costs (Costs Infra Sheet)The following costs items are accounted for:• Costs of cables.• Costs of civil engineering.• Costs of microwave systems.

The following table summarizes how these ele-ments are taken into account:

The calculations are similar to those of the otherCost sheets.

Total Costs (Tot Sheet)The Tot sheet consolidates all the data calculated in theprevious sections and calculates cost annuities for net-work elements used to provide interconnection ser-vices (see lines 34 to 40 of the Ucosts sheet). In doingso, one obtains for the switching and the transmissionelements, respectively, the total traffic these elementscarry, and therefore derives their costs. Dividing thetotal cost of elements of a network by the total trafficgives the unit cost per minute.

The Mobile Network Calculations

The mobile network is modeled with three types ofnode: MSC-MSC, MSC-BSC, and BSC-BTS. Threesheets are devoted to the intermediary calculations ofthe interconnection costs for originating and terminat-ing services on the mobile network.

Capacity of the Mobile Network (Capa Mob Sheet)The capacity of the mobile network is determinedaccording two factors: the total traffic and the peakhour traffic.The sizing exercise is conducted for eachnetwork element.

Cost of the Mobile Network (Costs Mob Sheet)The number of equipment required and the cost com-putations are carried out on the same sheet.The Costsheet is similar in structure to the one reviewed earlierfor the fixed network (number of equipment, invest-ments, annualized costs, operating costs).


58

RCU- LS- TS- ToCable : LS LS TS IS

Cable 6 fibers MetersCable 12 fibers MetersCable 24 fibers MetersCable 36 fibers MetersCable 48 fibers MetersCable 96 fibers MetersTrenchDuct (meters) - transport alone Urban

SuburbanRural

Duct (meters) - shared Urban

SuburbanRural

Wireless

Radio RCU- LS- TS- To CS-system LS LS TS IS TS

155 Mbps Number34 Mbps Number8 Mbps Number2 Mbps NumberAntenna equipmentWireless siteswith environmentTotal number ofpylons by dimension– Light Number– Medium Number– Heavy Number

Other related costs


59

The calculation is thus based on:• Exchanges equipment functions, features, number

of equipment, number of subscribers, and numberof installed TRXs.

• Microwave systems allowing connections of remoteequipment to the exchanges.

• Capacity of leased lines.• The number of staff operating the network.

Total Costs of the Mobile Network (Tot Mob Sheet)The structure of the sheet is identical to that for thefixed network, apart from the fact that no system ofrealignment is implemented.

Switching MCS BSC BTSEquipment NumberSites NumberSubscriber lines (variable) NumberTRX NumberPrimary digital block (2 Mbit/s port) Number of ports

Microwave MCS-MCS MCS-BSC BSC-BTSRadio system155 Mbps Number34 Mbps Number8 Mbps Number2 Mbps NumberAntenna equipmentMicrowave sites with environmentTotal number of pylons by dimension– rooftop Number– Light Number– Medium Number– Heavy Number

Leased links MCS-MCS MCS-BSC BSC-BTSUrban NumberInterurban NumberInterurban Km

MCS BSC BTSPersonnel Number

MCS-MCS MCS-MCS BSC-BTSPersonnel Number

60

Appendixes

Appendix 1: Economic Cost Approach

Calculation without Taking into Account Technical Progress The total discounted cost of using equipment for Nyears in the absence of technical progress is written:

where:• N is the economic lifetime of the investment; n the

current year.• I0 is the cost of the investment paid in year 0, or the

discounted sum in year 0 of the investment costs ifthese are spread over several years.

• fn is the operating-maintenance costs of year n, n =1, 2,…, N; these costs are generally increasing overtime,but generally the assumption is made that theyare constant and equal at f0.

• Vn is the resale value in year N (with conventionallyV0 = I0).We write ϕi

N (phi function) function1:

As defined above, the total discounted cost of theimplementation of the original investment I0 over theperiod N is equal to CN that is equivalent to: the dis-

counted sum of the original investment and of theoperating-maintenance costs less the discounted resalevalue of the equipment at N. Given the expressionretained for fn = f0, we have:

It is then necessary to consider the average annualeconomic cost of implementing this investment, whichcorresponds to the LRAIC.To do so the annual install-ment equivalent to the total discounted cost is considered(that is to say, the sum which, paid annually from year 1to year N, would enable it to be repaid).

The total annual cost is the sum of the:• total investment less the discounted residual value,

the whole discounted by phi (N,i) and

C If

i

V

iN

nn

NN

n

N

= ++( )

−+( )=

∑01 1 1

ϕNi n

N N

Ni

i

i i=

+= + −

+∑ 1

1

1 1

11 ( )

( )

( )

C If

i

V

i

I fi

V

i

I fV

i

N n

NN

N

n

NN

N

Ni

NN

= ++

−+

= ++

−+

= + −+( )

∑

∑

00

1

0 01

0 0

1 1

1

1 1

1

( ) ( )

( ) ( )

ϕ

XC

i

C If

V

i

IV

if

NN

n

NNN

iN

i

NN N

i

Ni

NN

=

+

= = + −+

= −+

+

∑ 1

1

1

1

1

1

1

00

0 0

( )

( )

(( )

)

ϕ ϕ ϕ

ϕ

• the annual operating costsThe sum XN spent every year during N years is

equivalent to the cost CN. It is equivalent to spend CNstraightaway or to spread expenditure XN over Nyears.

Calculation Taking into Account Technical ProgressWhen the technical progress is considered, there is aneed to introduce a factor capturing how the equip-ment is renewed over time (every N years).

This is the sum of a geometric series and can bewritten as:

If this expression is related to the constant annualinstallment equivalent to the cost CN, we have:

The value of the equipment during its lifetime isfrequently used in economic calculations.This is justi-fied by the need for a company to assess the valuationof the stock of equipment on a given date.What max-imum price would a company be willing to pay foridentical equipment (same age, same characteristics) ifthis equipment was lacking? In other terms,what is theopportunity cost that would be incurred by the loss ofthis equipment or the extra cost of anticipating itsrenewal by N* – n years where N* is its economic life-time and n its age (usage value or residual value). This lat-ter definition amounts to creating the fiction of aperfect secondhand market. In fact, in such a market,no businessperson would agree to buy used equipmentmore expensively than the net cost he/she would suf-fer if forced to buy new equipment prematurely, or sellequipment cheaper than the net cost it would entail toreplace it.

According to this definition, the residual value ofnew equipment on date n = 0 is its purchase price. Itsresidual value at date N* is nil, because the user will beindifferent to losing it. By assumption, the equipmentwould be replaced with identical equipment over atheoretically indefinite period. The maximum priceone is ready to pay for old equipment is determined bycomparing the costs obtained, either by procuringequipment of identical age or by buying new equip-ment straight away. More precisely, the usage value Unis such that the discounted costs corresponding to thetwo possible solutions are equal:• Payment of the price Un in year n, operating the

equipment in question from year n + 1 to year N,resale in N initially planned, renewal in identicalfashion over an indefinite period as from year N.

• Purchase in year n of new equipment, operatingduring N years, then renewal in identical fashionover an indefinite period.The usage value Un is then the maximum that the

company is ready to pay to pursue the operation withidentical equipment, at the same economic costprice as that obtained with the equipment in question.

In replacing the economic cost price by the sum ofthe operating costs fk and of the economic depreciationak (an = XN – fn), we have:

By subtracting the two consecutive years, we have:

otherwise,

an = (1 + i) • Un–1 – Un

where (1 + i)Un-1 is the discounted value in year n ofthe usage value Un-1.

Appendix 1: Economic Cost Approach

61

D CC

i

C

i

C

iN N

NN

NN

NkN

= ++

++

+ ++

+( ) ( )

...( )

...1 1 12

DC

i

i C

iN

N

N

NN

N=

−+( )

=+( )+( ) −1

1

1

1

1 1

DX

iNN=

U

i

f

i

V

i

X

i

nn

kk

k n

NN

NN

kk n

N

1 1 1 11 1+( )+

+( )−

+( )=

+( )= + = +∑ ∑ *

U

i

a

i

V

i

nn

kk

k n

NN

N1 1 11+( )

=+( )

++( )= +

∑

U

i

U

i

a

i

nn

nn

nn

−−

+( )−

+( )=

+( )1

11 1 1

The economic depreciation is thus interpreted as aloss of discounted usage value.2 The economic depre-ciation is thus an annual cost representing the useof the capital. More practically, this economic depre-ciation can be obtained by subtracting the amount ofoperating costs at each time n from the economic costprice.

Let us make simplifying assumptions. Suppose VN =0 and technical progress over time. If the cost of theinvestment decreases regularly at a rate g, then,

By h, we mean the composite rate defined by (1 +h) = (1 + i) * (1 + g).

Moreover, we retain operating costs constant overtime, fn = f0.

In the first case, the economic depreciation is equalto the constant annual installment equivalent to theinvestment costs. Technical progress comes down toincreasing the rate of discount i by a sum g, that is tosay, taking it as equal to h.

We then have:

since j = i (growth of operating costs nil).Thus, taking into account technical progress comes

down to increasing the rate of remuneration of capital

i by a factor g, which is the decrease in prices over theperiod in question.

Recurrent and Nonrecurrent Section of Debt RecoveryOnce an annual economic cost C for a given service isdetermined, it can be recovered over a lifetime T of theproductive resource. Assuming Rt, the recurrentamount recovered every year, and a nonrecurrent formR1 recoverable in year 1.

The principle that must prevail is that the dis-counted sum of the Rt’s increased with the nonrecur-rent part R1 be equal to the total C to be recovered,that is:

In general, one accepts recovery of the specificinterconnection costs (costs of co-localization, of con-nections between operators, and so forth) in the non-recurrent part and the costs of network use (callorigination and termination) in the recurrent part.

The model given here does not deal with specificcosts.


62

I I gnn= +0 1/ ( )

XI

fI

fN Nh

Nj

Ni

Nh

= + = +00

00ϕ

ϕϕ ϕ

C RR

it

tt

T

= ++=

∑11 1( )

Notes

1. Geometric sum of reason 1/(1 + i)If Sn = a + a2 + a3 + … + an, then we havethe following equations:Sn – Sn–1 = an and a + aSn–1 = Sn and thatgives:

by putting a = 1/(1 + i),we obtain 1 – a = ia and:

2. If a more financial interpretation is given,the economic depreciation represents the

annuity by means of which the initial invest-ment could be repaid if it were borrowed ata rate equal to the discount rate.The nomi-nal loss of usage value thus corresponds tothe share of repayment in capital of this an-nuity. The usage value thus corresponds tothe capital not yet repaid.

S aaan

n= −

−11

Si

i i i in

n

n n= + −

+= −

+( )

( )(

( ))

1 1

1

11

1

1

Appendix 2: Capital Cost Approach

The cost of the operators’ capital must reflect theopportunity cost of the funds invested in the compo-nents of the network and the other connected assets.Traditionally it reflects the following elements:• The average (weighted) cost of the indebtedness for

the various means of financing available to eachoperator.

• The cost of the equity capital, measured by thereturn the shareholders require, to invest in the net-work—taking into account the risks tied to thisinvestment.

• The value of the borrowed capital and the equitycapital.This information can then be used to determine

the weighted average cost of capital (WACC) accord-ing to the following formula:

WACC = re × E/(D + E) + rd × D/(D + E)

where re is the cost of the equity capital, rd is the cost ofborrowing,E is the total value of the equity capital, andD is the total value of the interest-producing debt.

The calculation of the WACC for a given operatorconsidered globally would be relatively direct, providedpossible arguments on the exact calculation and thevalue of the input data of the WACC formulas are setaside. Nevertheless, it may be that the regulators mustestablish whether the application of the global capitalcost represented by the WACC is appropriate for theregulated activities of the operators; when such is thecase, the global WACC could serve to determine theinterconnection fees.

Otherwise, the regulators can take into account thefact that various risk premiums are normally applicableto different activities, which could translate into differ-ences on the level of the cost of equity capital re, evenif the financial structure is the same. In that case, therecould be a different WACC for each branch of activityor each activity broken down (mobile telecommunica-tions, cable television, or international services).

The financial economy, and the actual behavior, ofinvestors teache that the cost of equity capital re is equalto the cost of borrowing without risk, to which isadded a risk premium that depends on the activity

invested upon, and on the financial market in demand.Activities where competition is liveliest usually entail ahigher risk. The cost of borrowing rd also variesbetween activities and between companies but to alesser extent than the cost of equity capital re for agiven financial market. As far as the structure of thecapital (E and D), is concerned, it should also reflectthe balance sheet of each main activity.When there isonly one main balance sheet for several activities, it isacceptable to assume that these activities share the samecapital structure. In this light, it can usually be assumedthat the cost of borrowing rd is the same for all activi-ties, unless their results are judiciously different.

The WACC must be applied to a capital value forthe components of the network and the other relatedassets, in order to determine the return to be attained,with the help of interconnection fees.While it is rela-tively easy to determine the value of borrowed capitaland equity capital for an operator as a whole, it is noteasy to determine these values for each of the opera-tor’s activities.This is because decisions on financing byborrowing are to a large extent company decisions.They are determined by various factors, such as thehistorical loan facilities and tax management consider-ations. It follows that the indebtedness of the companyis liable not to correspond exactly to the financialneeds of its various activities.

To fix prices, the regulators and the operators areinterested in the average capital employed during agiven period, rather than the capital employed at agiven moment, for example, at the end of the financialyear.This is justified by the fact that a “snapshot” of thesituation, at a given time, will not likely represent theaverage level of capital committed by the operator.Tobe precise, the balance of working capital at a giventime cannot represent the average need for liquid assetsover a lengthy period.The separate operators’ account-ing must, therefore, indicate the average capitalengaged, and not an end-of-year balance.

Accounting Values and Market ValuesThe great increase in the valuation of telecommunica-tions operators in 1999–2000 raised the question of thechoice of an accounting value for E (equity capitalentered in the balance sheet) or of a market value(stock market capitalization). It is generally accepted


63

that the cost of capital must reflect the minimumremuneration expected by the fund providers as awhole (shareholders and creditors). From this point ofview, the E value should reflect the stock market capi-talization as long as this is not the result of a speculativebubble. In fact, in this latter case, the financial marketsare inefficient.Therefore, regulators and operators mustagree on a sort of target financial structure where thecapitalization is coherent with the expectation offuture profits.

This assessment is important. In fact, the indebted-ness lever (D/E relation) can be very different depend-ing on whether an accounting value of E or a stockmarket value is considered; however, as re is greaterthan rd, this choice could exert a considerable impacton the WACC.

Effect of Profits TaxIn the WACC formula, re represents the cost requiredon the share capital and rd the cost of the financialdebt; this is situated before tax, that is to say that rd = rd(1 – θ), where rd is the actuarial rate of the financialdebt and q the rate of profits tax.

The use of WACC in determining tariffs is not incommon use and leads to adapting it. As it is definedabove, the WACC is net of tax. However, the tariff of aservice is fixed before tax. It is therefore necessary toincrease the WACC by the rate of profits tax and to usethe corrected WACC:

where rd* represents the actuarial rate of the financialdebt.

Appreciation of re and rdre and rd are calculated on the basis of a reference ratecalled rate without risk, corresponding in general tothe return on long-term (10-year) risk-free bonds(state bonds).To this rate a risk premium must be addedcorresponding to the activity involved and to the typeof finance considered.

The risk premium may be assessed on the past fluc-tuations of the security, or better on the fluctuations ofsimilar securities (sector assessment) so as to apprehend

what is called the market trend, linking the expectedprofitability to the level of risk.The return of the sharesis then expressed as follows:

where:rf is the risk-free return.rm is the average return expected on the market (for

example, the return represented by a market referenceindex of the type of Dow Jones, S&P, CAC) (level ofmarket return).

β is a weighting coefficient of the market differen-tial (equity beta).

It is generally accepted that interconnection repre-sents less risk than the fixed telephony activity, whichitself represents less risk than mobile telephony. As ingeneral a rate of return corresponding to the globalactivity is retained (for the reasons recalled at the startof this appendix), the risk premium tends to decreasewhen the interconnection activity increases.

If rf and rm are given by the financial market con-cerned, the regulator must assess the operator’s betacoefficient, and adapt it, to take into account the sharein its interconnection activity. β, therefore, measures anelasticity of the sensitivity of the operator’s security tovariations in the market index: if the profitability of themarket varies by 1 point, the profitability of the secu-rity varies by b points. In the African context, telecom-munications operators are generally considered as lessrisky values than the average value of the market: β istherefore generally lower than 1.

As far as rd* is concerned, it is generally calculatedas the risk-free return to which a premium (spread ordebt premium) specific to the operator concerned isadded.This return can be assessed either on the basis ofthe market values (risk-free return plus premium) oron the basis of contractual values presented by theoperator (annual average weighted cost of the opera-tor’s indebtedness)

Thus, the WACC taken into account in the modelis expressed as follows:


64

WACCE

D ED

D Ee

dr

r*( )

*=+

×+

+ ×+1 θ

r r r re f fm= + × −β ( )

WACCr r r E

D Er s

D

D E

f m ff p*

( )

( )( )= + × −

+×

++ + ×

+β

θ1


65

To calculate it, therefore, we need:• D/(D + E), share of the debt on the total financial

structure (which enables us to deduce E/(D + E);this ratio is sometimes called the level of gearing orleverage.

• rf, risk-free return of the financial market consid-ered.

• rm, average return of the financial market considered(expected growth of the reference stock marketindex).

• β, risk note of the security.• sp, risk premium of the operator.• θ , rate of profits tax.

These six values are required to calculate theWACC. By default, values are proposed. For moredetails on the application of these concepts, refer toAlexander and others (1999).

Appendix 3: Radio Concentrator Solutions

Today, there are only a small number of TDMA tech-nology radio concentrator manufacturers. In fact, thepurchase of Lucent TRT by SRT, the giving up of suchsolutions by the majors such as Alcatel or Siemens,means that SRT and NEC and a few small constructorsfrom low-density countries (Australia) remain the lastsuppliers of these technologies, which are now chal-lenged by other solutions (fixed global system mobile,satellite, and so forth).

Radio concentrators are particularly suitable forlow-density rural zones with wide gaps between eachvillage.They enable subscribers far from the exchange(up to 1,600 km) to be connected, and can tolerateconstraints specific to rural areas (for example, electricpower provided by solar panels). A leap between sta-tions can be up to 50 km.

These systems can operate in the 500 MHz, 1.5GHz, 2.5 GHz, and 3.5 GHz bands.The last leap canbe by wire or else wireless thanks to a wireless localloop termination.These systems offer basic telephoneand public phone services, group 3 fax, data transmis-sion services as well as basic Integrated system digitalnetwork (2B+D) services. They use TDMA distribu-tion, which enables the available spectrum to be opti-mized.The nodal and remote stations connected onlyuse one frequency pair.They often use Yagi antennas.

The capacity of these systems can go up to 4,096subscribers and traffic of 188 E. It should be noted thatthese calculations are carried out for a loss rate of 1percent. In fact, the traffic capacity is greater becauselocal calls do not use resources.Two subscribers con-nected to the same station can communicate withoutoccupying a channel.

The central station (CS) connects to the splitterof the automatic exchange of the public network.TheCS is linked by microwave to the remote stations.Theradio subassembly may be deported and linked by cableor radio at 2 Mbit/s.

The repeater station (RS) serves as a relaybetween sites not seen by the central station. Therepeater station can provide up to 120 circuits at 64kbit/s or 240 circuits at 32 kbit/s.

The terminal station (TS; SDE on the diagram)connects the subscribers.The terminal station can con-nect cabled subscribers or those with a DECT solution.


66

Figure A3.1 Example of the Architecture of an IRT(TRT-Lucent Network)

Table A3.1 Number of Subscribers Depending onthe Traffic per Subscriber (in mE)

Traffic per subscriber Number of subscribers

50 102460 85070 70080 65090 600

100 520110 500120 470130 450140 420150 400

Note: It is assumed that 30 percent of calls are local; otherwise, the systemloses between 5 percent and 15 percent of its subscribers.

67

BibliographyThe word processed describes informally re-produced works that may not be commonlyavailable through libraries.

Alexander, I.,A. Estache,A. Oliveri. 1999.“AFew Things Transport Regulators ShouldKnow about Risk and the Cost of Capi-tal.”The World Bank Institute. Processed.

Armstrong, M. 1998. “Network Intercon-nection in Telecommunications.” Eco-nomic Journal 108:545–64.

Armstrong, M., C. Doyle, and J. Vickers.1996. “The Access Pricing Problem: ASynthesis.” Journal of Industrial Economics44:131–50.

Armstrong,M., and J.Vickers.1998.“The Ac-cess Pricing Problem with Deregulation:A Note.” Journal of Industrial Economics46:115–21.

Benitez, D., A. Estache, M. Kennett, and C.Ruzzier. 2000.“Are Cost Models Usefulfor Telecoms Regulators in DevelopingCountries?” Policy Research WorkingPaper 2384.The World Bank Institute.

BIPE. 2000. “What Telecom Regulation forLow-Income African Countries?” Studyfor the European Commission. Availableat http://europa.eu.int/ISPO/intcoop/i_acp.html.

BIPE. 2003. Field Units Reports:Cameroon, Côte d’Ivoire, Zambia, Burk-ina Faso.

Celani, M. O. D. Petrecolla, and C. A.Ruzzier. 2002. “Desagregación de redesen telecomunicaciones. Una visión desdela política de defensa de la competencia.”Processed.

Economides, N. 1996. “The Economics ofNetworks.” International Journal of Indus-trial Organization 14:673–99.

———.1998.“The Incentive for Non-PriceDiscrimination by an Input Monopolist.”International Journal of Industrial Organiza-tion 16:271–84.

Economides,N., and L. J.White.1995.“Accessand Interconnection Pricing: How Effi-cient Is the ‘Efficient Component PricingRule’?” Antitrust Bulletin 40:557–79.

Hausman, J. 1996. “Reply Affidavit of Prof.Jerry Hausman, FCC CC Docket No.96-98” and “Valuation and the Effect ofRegulation on New Services inTelecommunications.” Brookings Paperson Economic Activity: Microeconomics.

Jamison, M. 1998. “Regulatory Techniquesfor Addressing Interconnection, Access,and Cross-Subsidy in Telecommunica-tions.” In Margaret Arblaster and Mark A.Jamison, eds., Infrastructure Regulation andMarket Reform: Principles and Practice. Aus-tralian Competition and ConsumerCommission and the Public Utility Re-search.

———. 1999.“Does Practice Follow Princi-ple? Applying Real Options Principles ToProxy Costs in U.S. Telecommunica-tions.” In James Alleman and Eli Noam,eds., Real Options: The New InvestmentTheory and its Implications for Telecommuni-cations Economics. Boston: Kluwer Acade-mic Publishers.

Laffont, J. J. 1998.“Translating Principles intoPractice.” EDI Regulatory Reform Dis-cussion Paper. World Bank, Washington,D.C. http://www.worldbank.org/wbi/regulation/wp.htm.

Laffont, J. J., and J. Tirole. 1996. “CreatingCompetition through Interconnection:Theory and Practice.” Journal of Regula-tory Economics 10:227–56.

———. 2001. Competition in Telecommunica-tions. Munich Lectures, The MIT Press,2nd ed.

Laffont, J. J., P. Rey, and J.Tirole. 1998.“Net-work Competition: I. Overview andNon-discriminatory Pricing; II. Discrim-inatory Pricing.” RAND Journal of Eco-nomics 29:1–56.

Salinger, M. 1998. “Regulating Prices toEqual Forward Looking Costs: Cost-Based Prices or Price-Based Costs?” Jour-nal of Regulatory Economics 14:149–64.

Sidak, J. G., and D. Spulber. 1997. “TheTragedy of the Telecommons: Govern-ment Pricing of Unbundled Network El-ements under the TelecommunicationsAct of 1996.” Columbia Law Review71:1081–161.

Spulber, D., and J. G. Sidak. 1997.“NetworkAccess Pricing and Deregulation.” Indus-try and Corporate Change 6:757–82.

Valletti,T.1998.“The Practice of Access Pric-ing: Telecoms in the UK.” The WorldBank, EDIRP. Processed.

———.2001.“The Theory of Access Pricingand Its Linkage with Investment Incen-tives.” In Martin Cave, Papers on AccessPricing Investment and Entry inTelecommunications. Warwick BusinessSchool, Research paper series.

Valletti,T., and A.Estache.1998.“The Theoryof Access Pricing:An Overview for Infra-structure Regulators.” The World BankInstitute.Available at http://www.world-bank.org/wbi/regulation/wp.htm.

Wildman, S. S. 1997.“Interconnection Pric-ing, Stranded Costs, and the OptimalRegulatory Contract.” Industrial and Cor-porate Change 6:741–55.

A Model for CalculatingInterconnection Costs inTelecommunications

1818 H Street, NW


U.S.A.

Telephone: 202.473.1000

Fax: 202.477.6391

Web Site: www.worldbank.org


PPIAF Program Management Unit

c/o The World Bank

1818 H Street, NW, MSN I9-907


U.S.A.

Telephone: 202.458.5588

Fax: 202.522.7466

Web Site: www.ppiaf.org


™xHSKIMBy356715zv,:&:':=:"ISBN 0-8213-5671-2

THE WORLD BANK

PPIAF

PPIAF

THE WORLD BANK

A M

od

el fo

r Calc

ula

ting

Inte

rco

nn

ectio

n C

osts

in T

ele

co

mm

un

icatio

ns

TH

E W

OR

LD

BA

NK

The liberalization of the telecommunications markets in Sub-Saharan Africa led toincreased competition on the provision and pricing of communication services. But,

due to the lack of appropriate regulatory tools, newly established regulators are poorlyequipped to arbitrate increasing interconnection disputes between competing operators.This guidebook and its associated CD-ROM, including the cost model, were prepared toprovide Sub-Saharan Africa regulators and operators with a sound regulatory tool allowingthe determination of accurate interconnection costs, thus facilitating the settlement oflengthy and costly interconnection disputes between fixed and mobile operators. The costmodel belongs to the family of “Bottom-Up” models, which calculate interconnectioncost incurred by an efficient operator using the Long Run Incremental Cost (LRIC)methodology. The proposed cost model takes into account most features characterizingthe development stage of telecommunications networks in Sub-Saharan Africa (smallsize of fixed network, importance of rural telephony, excessive reliance on microwavetechnology, explosive demand for mobile service, and weak regulatory capacity).

Paul Noumba Um

Laurent Gille

Lucile Simon

Christophe Rudelle

a model for calculating interconnection costs in ... · a model for calculating interconnection...

Documents